From 5a3ff06d9fcf555cb3ca99675b94dc0ac5787d32 Mon Sep 17 00:00:00 2001 From: Waleed Latif Date: Thu, 19 Nov 2026 19:51:00 -0800 Subject: [PATCH 1/3] feat(comparisons): add Sim vs Dify and Sim vs Relevance AI comparison pages --- apps/sim/app/(landing)/comparisons/page.tsx | 7 +- apps/sim/app/(landing)/comparisons/utils.ts | 4 + apps/sim/components/icons.tsx | 34 + apps/sim/lib/compare/data/competitors/dify.ts | 1360 ++++++++++++++++ .../compare/data/competitors/relevance-ai.ts | 1416 +++++++++++++++++ apps/sim/lib/compare/data/index.ts | 2 + apps/sim/scripts/test-landing-seo-e2e.ts | 4 + 7 files changed, 2824 insertions(+), 3 deletions(-) create mode 100644 apps/sim/lib/compare/data/competitors/dify.ts create mode 100644 apps/sim/lib/compare/data/competitors/relevance-ai.ts diff --git a/apps/sim/app/(landing)/comparisons/page.tsx b/apps/sim/app/(landing)/comparisons/page.tsx index 5f62a634c09..9d88134e7db 100644 --- a/apps/sim/app/(landing)/comparisons/page.tsx +++ b/apps/sim/app/(landing)/comparisons/page.tsx @@ -120,9 +120,10 @@ export default function ComparisonHubPage() {

This directory lists every Sim vs. competitor comparison page, covering workflow automation platforms (n8n, Zapier, Make, Pipedream), enterprise AI builders (Gumloop, - Workato, Retool, Tines, StackAI, Power Automate, Vellum), and AI agent products - (OpenAI AgentKit, Claude Cowork). Each page gives sourced, dated facts across - platform, AI capabilities, integrations, pricing, security, and support. + Workato, Retool, Tines, StackAI, Relevance AI, Power Automate, Vellum), open-source + agent builders (Dify, Langflow, Flowise), and AI agent products (OpenAI AgentKit, + Claude Cowork). Each page gives sourced, dated facts across platform, AI capabilities, + integrations, pricing, security, and support.

diff --git a/apps/sim/app/(landing)/comparisons/utils.ts b/apps/sim/app/(landing)/comparisons/utils.ts index 7fa3fc55078..a9784261a4d 100644 --- a/apps/sim/app/(landing)/comparisons/utils.ts +++ b/apps/sim/app/(landing)/comparisons/utils.ts @@ -3,6 +3,7 @@ import { type CompetitorProfile, claudeCoworkProfile, crewaiProfile, + difyProfile, dustProfile, flowiseProfile, gumloopProfile, @@ -16,6 +17,7 @@ import { type Prose, pipedreamProfile, powerAutomateProfile, + relevanceAiProfile, retoolProfile, simProfile, stackaiProfile, @@ -42,11 +44,13 @@ export const ALL_COMPETITORS: CompetitorProfile[] = [ openaiAgentkitProfile, tinesProfile, stackaiProfile, + relevanceAiProfile, powerAutomateProfile, vellumProfile, claudeCoworkProfile, langflowProfile, flowiseProfile, + difyProfile, microsoftCopilotProfile, openClawProfile, dustProfile, diff --git a/apps/sim/components/icons.tsx b/apps/sim/components/icons.tsx index 190e8091440..05b17279bbd 100644 --- a/apps/sim/components/icons.tsx +++ b/apps/sim/components/icons.tsx @@ -1828,6 +1828,40 @@ export function DustIcon(props: SVGProps) { ) } +export function DifyIcon(props: SVGProps) { + return ( + + + + + ) +} + +export function RelevanceAIIcon(props: SVGProps) { + return ( + + + + + + ) +} + export function OpenClawIcon(props: SVGProps) { const id = useId() const gid = (name: string) => `openclaw_${name}_${id}` diff --git a/apps/sim/lib/compare/data/competitors/dify.ts b/apps/sim/lib/compare/data/competitors/dify.ts new file mode 100644 index 00000000000..dc84b05ab31 --- /dev/null +++ b/apps/sim/lib/compare/data/competitors/dify.ts @@ -0,0 +1,1360 @@ +import { DifyIcon } from '@/components/icons' +import type { CompetitorProfile } from '@/lib/compare/data/types' + +/** Researched and cross-verified against live vendor sources on 2026-11-19. */ +export const difyProfile: CompetitorProfile = { + id: 'dify', + name: 'Dify', + website: 'https://dify.ai', + brand: { + icon: DifyIcon, + selfFramed: false, + colors: ['#0033FF', '#000000'], + source: 'Dify logo and favicon SVGs (langgenius/dify web/public/logo, dify.ai/favicon.svg)', + asOf: '2026-11-19', + }, + oneLiner: + 'Dify is an LLM application platform from LangGenius for building workflows, chatflows, agents, and retrieval-augmented generation (RAG) knowledge bases in a visual editor, offered as Dify Cloud, a self-hosted Community Edition under a modified Apache 2.0 license, and a commercial Enterprise edition.', + leadAnswer: [ + 'Dify is an LLM application platform for building chat apps, workflows, agents, and retrieval-augmented generation (RAG) knowledge bases, available as Dify Cloud, a self-hosted Community Edition, and a commercial Enterprise edition. Its source is published under the ', + { + text: 'Dify Open Source License', + href: 'https://github.com/langgenius/dify/blob/main/LICENSE', + }, + ', a modified Apache 2.0 license that requires a commercial license to run a multi-tenant service or to remove the Dify logo from its frontend. Sim is the open-source AI workspace, with a standard Apache 2.0 core, for building, deploying, and managing AI agents visually, conversationally, or with code. Choose Dify when your product centers on a managed knowledge base and a large plugin marketplace. Choose Sim when you need an unmodified Apache 2.0 core, deployed agents connected to the business apps your team already uses, and published production controls such as failure alerts and execution limits on its cloud plans.', + ], + betterThanAnswer: [ + "Sim is the stronger fit when AI agents have to act across business systems and you want a standard Apache 2.0 core you can self-host without the multi-tenant condition that Dify's license adds. Dify is the stronger fit when the deliverable is a chat assistant or knowledge app built on a deep retrieval stack, with a large community plugin marketplace behind it. Both have visual builders, human approval steps, loops, parallel branches, and Model Context Protocol (MCP) support for connecting external tools, so the license terms and the shape of the work usually decide it.", + ], + sectionIntros: { + platform: [ + 'Sim and Dify both offer a visual builder, self-hosting, version history, and live co-editing. They differ on license terms, file storage and data tables, and how far environment promotion goes.', + ], + pricing: [ + 'Sim charges per user with usage credits and lets you bring your own model keys. Dify Cloud charges per workspace with monthly message credits, and its paid plans also accept your own model keys.', + ], + security: [ + 'Dify publishes SOC 2 and ISO 27001 badges and reserves SSO, SCIM, audit logs, and credential governance for its Enterprise edition. Sim documents SOC 2, and its Enterprise plan adds SSO, SCIM, an audit log API, and model, tool, and credential controls.', + ], + aiCapabilities: [ + 'Both products ship agents, a knowledge base, Model Context Protocol (MCP) client and server support, human input steps, loops, and parallel branches. Dify adds sandboxed agents and a skills library, while Sim adds an evaluator block, PII masking, and agent-to-agent (A2A) support.', + ], + integrations: [ + "Dify extends through more than 1,000 Marketplace plugins built with its plugin SDK, plus OpenAPI imports and MCP servers. Sim's integrations are first-party blocks, extended through custom code, the REST API, and MCP.", + ], + observability: [ + 'Both trace runs step by step, retry failed steps, route failures to an error branch, and export traces to external tools. Sim also documents failure alerts and cloud execution limits that Dify does not publish for Dify Cloud.', + ], + support: [ + "Both offer community channels and documentation. Dify's paid support runs from priority email on Cloud to a named customer success manager on Enterprise, while Sim offers community support plus Enterprise dedicated support.", + ], + }, + faqs: [ + { + question: 'Is Dify really open source, compared to Sim?', + answer: + "Dify's source code is public, but its Dify Open Source License is a modified Apache 2.0 license. It requires a commercial license to use the code to operate a multi-tenant environment, where one tenant is one workspace, and forbids removing or modifying the Dify logo or copyright information in the Dify console or applications. Dify's own homepage describes the Community Edition as source-available under an Apache-2.0-derivative license. Sim's core uses the standard Apache 2.0 license, and only its enterprise features sit under a separate Sim Enterprise License.", + }, + { + question: 'Which is better for RAG and knowledge bases, Sim or Dify?', + answer: + "Dify is the more specialized choice when the knowledge base is the product. It offers classic knowledge bases with vector, full-text, and hybrid search plus reranking, customizable Knowledge Pipelines, connections to external knowledge bases, and a beta Agent Knowledge Base that organizes documents with a reasoning model. Sim's knowledge base combines vector and full-text search over 15 file types with synced connectors, and fits teams that use retrieval as one step inside a larger agent workflow that also updates other systems.", + }, + { + question: 'How do Sim and Dify compare on pricing?', + answer: + "Dify Cloud prices per workspace: Sandbox is free with 200 message credits, Professional is $59 per workspace per month ($590 per year) with 3 members and 5,000 message credits a month, and Team is $159 per workspace per month ($1,590 per year) with 50 members and 10,000 message credits a month. Sim's Pro plan is $25 per user per month with usage credits, and its Free plan includes 1,000 one-time credits. Both let you use your own model provider keys, and Dify Enterprise pricing is custom.", + }, + { + question: 'Can I self-host Dify or Sim for many internal teams?', + answer: + "Sim's core uses the standard Apache 2.0 license, which has no multi-tenant condition, and self-hosts with Docker Compose or Kubernetes; SSO and audit logs require an Enterprise license. Dify's Community Edition self-hosts with Docker Compose and is limited to a single workspace; running multiple workspaces requires Dify Enterprise, because the Dify Open Source License requires written authorization to operate a multi-tenant environment, and Enterprise adds Kubernetes deployment with high availability.", + }, + ], + standoutFeatures: [ + { + title: 'A deep retrieval stack for knowledge-centered apps', + description: + 'Classic knowledge bases offer vector, full-text, and hybrid search with optional reranking, metadata filtering, and a Retrieval Testing page that shows which chunks come back. Knowledge Pipelines let you design your own extraction and chunking steps, external knowledge bases can be searched through an API without copying content in, and a beta Agent Knowledge Base uses a reasoning model to organize documents and supports golden-question tests.', + shortDescription: + 'Hybrid search, reranking, custom knowledge pipelines, and an agent-organized knowledge base.', + source: { + url: 'https://docs.dify.ai/en/cloud/use-dify/knowledge/create-knowledge/setting-indexing-methods', + label: 'Dify Docs: Specify the Index Method and Retrieval Settings', + asOf: '2026-11-19', + }, + }, + { + title: 'A large plugin Marketplace with a public plugin SDK', + description: + "Dify Marketplace lists more than 1,000 plugins across tools, model providers, agent strategies, extensions, data sources, and triggers, from Dify, partners, and community developers. Dify says each Marketplace plugin is reviewed and tested before publication, and anyone can build private or public plugins with Dify's plugin SDK and CLI.", + shortDescription: 'More than 1,000 reviewed plugins, and an SDK to build your own.', + source: { + url: 'https://docs.dify.ai/en/cloud/use-dify/workspace/plugins', + label: 'Dify Docs: Integrations', + asOf: '2026-11-19', + }, + }, + { + title: 'Sandboxed agents built by chatting, with a shared skills library', + description: + "Dify's new Agent (beta) runs in its own sandbox with 20 GB of storage, where it runs commands, installs programs, and reads and writes files. You can build it by describing what it should do, give it skills in the open SKILL.md format from a workspace library of up to 500 skills, and use it as a chat app or as a step inside a workflow.", + shortDescription: + 'Beta agents with their own sandbox, built by chatting, using shared skills.', + source: { + url: 'https://docs.dify.ai/en/cloud/use-dify/build/new-agent/build', + label: 'Dify Docs: Build an Agent', + asOf: '2026-11-19', + }, + }, + { + title: 'Human approvals delivered to email and team chat apps', + description: + 'The Human Input node pauses a workflow and sends a form with custom decision buttons through the web app or email, with a configurable timeout branch. Since version 1.20.0, requests can also go to Slack, Feishu/Lark, Microsoft Teams, DingTalk, or WeCom, using a shared contact list and notification templates.', + shortDescription: 'Approval forms by web app, email, Slack, Teams, and other chat apps.', + source: { + url: 'https://github.com/langgenius/dify/releases/tag/1.20.0', + label: 'Dify v1.20.0 release notes (GitHub)', + asOf: '2026-11-19', + }, + }, + ], + limitations: [ + { + title: 'Modified Apache 2.0 license with multi-tenant and branding conditions', + description: + 'The Dify Open Source License requires a commercial license to use the source code to operate a multi-tenant environment (one tenant equals one workspace) without written authorization, and forbids removing or modifying the Dify logo or copyright information in the Dify console or applications. Contributors must also agree that the producer can make the license more strict or more relaxed.', + shortDescription: + 'Multi-workspace hosting and logo removal need a commercial license from Dify.', + source: { + url: 'https://github.com/langgenius/dify/blob/main/LICENSE', + label: 'langgenius/dify LICENSE (GitHub)', + asOf: '2026-11-19', + }, + }, + { + title: 'Self-hosted Community Edition is limited to a single workspace', + description: + "Dify's pricing lists the free Community Edition as a single workspace without SSO, role management, or high availability. Multiple workspaces, SSO, SCIM, audit logs, credential governance, and Kubernetes high availability require the commercial Enterprise edition, which has no published price.", + shortDescription: + 'Community self-hosting is one workspace; SSO and multi-workspace need Enterprise.', + source: { + url: 'https://dify.ai/pricing/dify-enterprise', + label: 'Dify Pricing: Self-Hosted and Enterprise plan comparison', + asOf: '2026-11-19', + }, + }, + { + title: 'Code steps run on a fixed package set with tight output limits', + description: + "Code nodes run Python or JavaScript in Dify's sandbox with a standard set of preinstalled packages, and imports outside that set fail at run time. Outputs are capped (for example, 500 items per array and 400,000 characters per string), and the self-hosted sandbox defaults to 15 seconds per code run.", + shortDescription: + 'Code nodes allow only preinstalled packages and cap output size and run time.', + source: { + url: 'https://docs.dify.ai/en/cloud/use-dify/nodes/code', + label: 'Dify Docs: Code node', + asOf: '2026-11-19', + }, + }, + ], + facts: { + platform: { + builderType: { + value: + 'Dify builds Workflow and Chatflow apps on a shared visual node editor where most work is drag, connect, and configure, plus simpler Chatbot, Text Generator, and Agent apps configured through forms. Agents can also be built by chatting, and the difyctl CLI and OpenAPI let coding agents build and run apps from a terminal.', + detail: + 'A Workflow runs once from input to output; a Chatflow adds a conversation layer where each message runs the flow.', + shortValue: 'Visual node editor plus form and chat-built agents', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/build/workflow-chatflow', + label: 'Dify Docs: Workflow & Chatflow', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/cli/overview', + label: 'Dify Docs: difyctl overview', + asOf: '2026-11-19', + }, + ], + }, + learningCurve: { + value: + "Low for chat apps and templates, moderate for workflows: Dify's enterprise customers describe it as accessible to beginners, and it publishes a 10-lesson Workflow 101 tutorial. Tuning retrieval settings, variables, and self-hosted deployments still requires technical knowledge.", + shortValue: 'Low for chat apps, moderate for workflows', + confidence: 'estimated', + sources: [ + { + url: 'https://docs.dify.ai/en/learn/tutorials/workflow-101/lesson-01', + label: 'Dify Docs: Workflow 101, Lesson 1', + asOf: '2026-11-19', + }, + { + url: 'https://dify.ai/dify-enterprise', + label: 'Dify Enterprise page (customer quotes)', + asOf: '2026-11-19', + }, + ], + }, + selfHostOption: { + value: + 'Yes: the Community Edition self-hosts with Docker Compose (2 CPU cores and 4 GiB RAM minimum) or from source, subject to the Dify Open Source License, and is limited to a single workspace. Dify Enterprise adds Kubernetes deployment with a Helm chart, high availability, and multiple workspaces.', + shortValue: 'Yes, Docker Compose; Kubernetes via Enterprise', + confidence: 'verified', + sources: [ + { + url: 'https://github.com/langgenius/dify', + label: 'GitHub: langgenius/dify README (Quick start)', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/self-host/deploy/overview', + label: 'Dify Docs: Deploy Dify', + asOf: '2026-11-19', + }, + { + url: 'https://dify.ai/pricing/dify-enterprise', + label: 'Dify Pricing: Self-Hosted and Enterprise plan comparison', + asOf: '2026-11-19', + }, + ], + }, + deploymentOptions: { + value: + 'Dify Cloud (managed SaaS), the self-hosted Community Edition (Docker Compose or source), Dify Premium on AWS, and Dify Enterprise, which Dify says deploys in your own VPC, on premises, or in an air-gapped network using a Helm chart for Kubernetes and Terraform modules for AWS, GCP, and Azure.', + shortValue: 'Cloud, Docker self-host, AWS, Enterprise VPC/air-gapped', + confidence: 'verified', + sources: [ + { + url: 'https://dify.ai/dify-enterprise', + label: 'Dify Enterprise page', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/self-host/deploy/platform-guides/dify-premium', + label: 'Dify Docs: Dify Premium on AWS', + asOf: '2026-11-19', + }, + { + url: 'https://enterprise-docs.dify.ai/en/3.13.x/deploy/advanced-configuration/dify-helm-chart', + label: 'Dify Enterprise Docs: Dify Helm Chart', + asOf: '2026-11-19', + }, + ], + }, + templates: { + value: + 'Yes: an Explore page of scenario-based app templates is included on every plan, agents can be created from templates you can try first, and Dify Marketplace also lists community workflow templates.', + shortValue: 'Yes, Explore templates plus Marketplace templates', + confidence: 'verified', + sources: [ + { + url: 'https://dify.ai/pricing/dify-cloud', + label: 'Dify Cloud plan comparison (Explore Page Templates)', + asOf: '2026-11-19', + }, + { + url: 'https://marketplace.dify.ai', + label: 'Dify Marketplace', + asOf: '2026-11-19', + }, + ], + }, + license: { + value: + 'Dify Open Source License: a modified Apache 2.0 license. It permits commercial use, but a commercial license is required to use the source code to operate a multi-tenant environment (one tenant equals one workspace) without written authorization, or to remove or modify the logo or copyright information in the Dify console or applications. Dify describes the Community Edition as source-available.', + detail: + 'The LICENSE file also says contributors agree the producer can make the license more strict or more relaxed, and that contributed code may be used commercially, including for Dify Cloud. The logo condition does not apply to uses that do not involve the web frontend.', + shortValue: 'Modified Apache 2.0 with multi-tenant and logo conditions', + confidence: 'verified', + sources: [ + { + url: 'https://github.com/langgenius/dify/blob/main/LICENSE', + label: 'langgenius/dify LICENSE (GitHub)', + asOf: '2026-11-19', + }, + { + url: 'https://dify.ai', + label: 'Dify homepage (Community Edition: source-available)', + asOf: '2026-11-19', + }, + ], + }, + environmentPromotion: { + value: + 'Partial: Dify Enterprise has an Environments feature, for Kubernetes deployments only, that deploys published app versions into separate runtime environments and flags when an environment is behind the latest version. Moving apps between workspaces otherwise means exporting and importing app packages, which carry inline agents, skills, and workflow tools.', + detail: + 'No documented workspace-level fork-and-promote flow with credential remapping across dev, QA, and prod.', + shortValue: 'Enterprise runtime environments; package export/import', + confidence: 'verified', + sources: [ + { + url: 'https://enterprise-docs.dify.ai/en/3.13.x/administer/environments', + label: 'Dify Enterprise Docs: Environments', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/cloud/use-dify/workspace/export-import-apps', + label: 'Dify Docs: Export and Import Apps', + asOf: '2026-11-19', + }, + ], + }, + versionControlDepth: { + value: + 'Yes: Workflow and Chatflow apps keep a working draft, the live version, and earlier published versions you can name, annotate with release notes, delete, and restore into the draft. Restoring a version, exporting a specific version, and running a specific version through the API require a paid plan.', + detail: + 'No diff or compare view or branching is documented. The Dify Cloud plan comparison still lists App Version Control as Coming Soon for Professional and Team, alongside the documented publish history.', + shortValue: 'Published version history with restore on paid plans', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/build/version-control', + label: 'Dify Docs: Version Control', + asOf: '2026-11-19', + }, + ], + }, + realtimeCollaboration: { + value: + 'Yes: workspace members with editor permissions can edit the same workflow at the same time with visible collaborator cursors, and leave in-editor comments with @mentions. When two members change the same element at once, the last edit wins.', + detail: + 'Documented in the self-hosted docs, where it runs over a separate WebSocket service enabled by ENABLE_COLLABORATION_MODE in the bundled Docker Compose stack.', + shortValue: 'Yes, live co-editing and comments; last edit wins', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/self-host/use-dify/build/workflow-collaboration', + label: 'Dify Docs: Collaborate with Teammates', + asOf: '2026-11-19', + }, + ], + }, + nativeFileStorage: { + value: + 'No: Dify handles files per app, knowledge base, and agent (uploads, agent files, and a 20 GB agent sandbox), but documents no workspace file system with folders, link-based sharing, or deleted-item recovery.', + shortValue: 'No workspace file storage with folders or sharing', + confidence: 'estimated', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/build/new-agent/build', + label: 'Dify Docs: Build an Agent (files and sandbox)', + asOf: '2026-11-19', + }, + ], + }, + dataTables: { + value: + "No: Dify's documentation describes no native spreadsheet-like data table. Structured data moves between nodes as variables, and knowledge bases store documents and chunks rather than editable rows.", + shortValue: 'No native spreadsheet-style data table', + confidence: 'estimated', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/knowledge/readme', + label: 'Dify Docs: Classic Knowledge Base', + asOf: '2026-11-19', + }, + ], + }, + richTextEditor: { + value: + 'Unknown: no public documentation describes an inline rich-text or WYSIWYG editor for documents stored in Dify. Skills edit text files in place, and Human Input forms render Markdown.', + shortValue: 'Unknown, no WYSIWYG document editor documented', + confidence: 'unknown', + sources: [], + }, + subWorkflows: { + value: + 'Yes: any Workflow that starts with a User Input node can be published as a workflow tool and called from a Tool node in another app, nested up to 5 levels deep by default on self-hosted deployments, with nested steps visible in run history and traces. Snippets also save groups of nodes for reuse, inserted as independent copies.', + shortValue: 'Yes, workflows published as tools, plus snippets', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/workspace/tools', + label: 'Dify Docs: Dify Tools (Workflow as Tool)', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/cloud/use-dify/build/snippet', + label: 'Dify Docs: Snippets', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/self-host/deploy/configuration/environments', + label: 'Dify Docs: Environment Variables (WORKFLOW_CALL_MAX_DEPTH)', + asOf: '2026-11-19', + }, + ], + }, + customBlocks: { + value: + 'Partial: Workflow as Tool publishes a workflow as a named tool with its own inputs that every app in the workspace can add, and invited agents pick up updates when their source agent is republished. Snippets are shared workspace-wide but insert as copies that do not follow later edits.', + detail: + 'Reuse is scoped to one workspace; no org-wide block catalog across workspaces with hidden internals and credentials is documented.', + shortValue: 'Partial, workflows as workspace tools; snippets copy', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/workspace/tools', + label: 'Dify Docs: Dify Tools (Workflow as Tool)', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/cloud/use-dify/nodes/agent', + label: 'Dify Docs: Agent node (Invite an Agent)', + asOf: '2026-11-19', + }, + ], + }, + }, + aiCapabilities: { + multiLlmSupport: { + value: + 'Dify supports hundreds of proprietary and open-source models from dozens of inference providers and self-hosted options, including any OpenAI API-compatible model, installed as model provider plugins (192 listed in the Marketplace). Dify Cloud also includes message credits for models from OpenAI, Anthropic, Gemini, xAI, and Tongyi.', + shortValue: 'Hundreds of models via provider plugins, incl. local', + confidence: 'verified', + sources: [ + { + url: 'https://github.com/langgenius/dify', + label: 'GitHub: langgenius/dify README (Comprehensive model support)', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/cloud/use-dify/workspace/model-providers', + label: 'Dify Docs: Model Providers', + asOf: '2026-11-19', + }, + { + url: 'https://marketplace.dify.ai', + label: 'Dify Marketplace (model plugins)', + asOf: '2026-11-19', + }, + ], + }, + agentReasoningBlocks: { + value: + 'Yes: the classic Agent node lets a model choose and call configured tools with Function Calling or ReAct strategies (more strategies install from the Marketplace), and the beta new Agent node runs a sandboxed agent as a workflow step that can pick its own output route. LLM, Question Classifier, and Parameter Extractor nodes cover single model calls.', + shortValue: 'Classic and sandboxed Agent nodes, plus LLM nodes', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/nodes/agent', + label: 'Dify Docs: Agent node', + asOf: '2026-11-19', + }, + ], + }, + naturalLanguageBuilding: { + value: + 'Partial: the new Agent (beta) can be built by chatting, where it writes its own prompt, skills, files, and environment variables, and a Skill Builder drafts skills from a description. Workflow and Chatflow graphs are built in the visual editor, or by a coding agent through the difyctl CLI.', + shortValue: 'Agents and skills built by chatting; workflows visual', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/build/new-agent/build', + label: 'Dify Docs: Build an Agent (Build by Chatting)', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/self-host/use-dify/build/skills', + label: 'Dify Docs: Skills (Skill Builder)', + asOf: '2026-11-19', + }, + ], + }, + knowledgeBaseRag: { + value: + 'Yes: classic knowledge bases import files, Notion, and websites, with configurable chunking, High Quality (vector, full-text, or hybrid search with optional reranking) or Economical indexing, metadata filtering, custom Knowledge Pipelines, and connections to external knowledge bases. A beta Agent Knowledge Base uses a reasoning model to organize documents and can search more deeply on request.', + detail: + 'Dify Cloud caps knowledge documents and storage per plan, from 50 documents and 50 MB on Sandbox to 1,000 documents and 20 GB on Team.', + shortValue: 'Yes, hybrid search, reranking, custom pipelines', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/knowledge/create-knowledge/setting-indexing-methods', + label: 'Dify Docs: Specify the Index Method and Retrieval Settings', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/cloud/use-dify/knowledge-base/overview', + label: 'Dify Docs: Agent Knowledge Base', + asOf: '2026-11-19', + }, + { + url: 'https://dify.ai/pricing', + label: 'Dify Pricing', + asOf: '2026-11-19', + }, + ], + }, + mcpSupport: { + value: + 'Yes: Dify connects to external Model Context Protocol (MCP) servers over HTTP transport and imports their tools for workflows and agents, with OAuth dynamic client registration and custom headers. Any Dify app can also be exposed as an MCP server.', + detail: 'Only MCP servers with HTTP transport are supported as clients.', + shortValue: 'Both MCP client (HTTP only) and MCP server', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/workspace/tools', + label: 'Dify Docs: Dify Tools (MCP)', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/cloud/use-dify/publish/publish-mcp', + label: 'Dify Docs: MCP Server', + asOf: '2026-11-19', + }, + ], + }, + evaluationGuardrails: { + value: + 'Partial: chat apps can filter inputs and outputs with OpenAI Moderation, a keyword list, or a custom moderation endpoint, and Agent Knowledge Bases save golden questions that must keep passing after changes. No general-purpose evaluator or scoring node for workflow outputs is documented.', + shortValue: 'Content moderation and retrieval tests; no evaluator node', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/build/additional-features', + label: 'Dify Docs: App Toolkit (Content Moderation)', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/cloud/use-dify/knowledge-base/overview', + label: 'Dify Docs: Agent Knowledge Base', + asOf: '2026-11-19', + }, + ], + }, + humanInTheLoop: { + value: + 'Yes: the Human Input node pauses a workflow and delivers a Markdown form with input fields and decision buttons through the web app or email, routes each button to its own branch, and follows a timeout branch after a configurable wait (3 days by default). Since 1.20.0, requests can also go to Slack, Feishu/Lark, Microsoft Teams, DingTalk, or WeCom, and since 1.17.0 the node works inside loops and iterations.', + shortValue: 'Human Input node via web app, email, Slack, Teams', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/nodes/human-input', + label: 'Dify Docs: Human Input node', + asOf: '2026-11-19', + }, + { + url: 'https://github.com/langgenius/dify/releases/tag/1.20.0', + label: 'Dify v1.20.0 release notes (GitHub)', + asOf: '2026-11-19', + }, + ], + }, + generativeMedia: { + value: + "Partial: text-to-speech and speech-to-text run on the workspace default speech models, and LLM nodes accept images for vision models. Image generation comes from Marketplace tool plugins rather than a built-in node, as in Dify's own image generation tutorial.", + shortValue: 'Speech built in; image generation via plugins', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/build/additional-features', + label: 'Dify Docs: App Toolkit (Text to Speech, Speech to Text)', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/learn/tutorials/build-ai-image-generation-app', + label: 'Dify Docs: AI Image Generation App tutorial', + asOf: '2026-11-19', + }, + ], + }, + dynamicToolUse: { + value: + 'Partial: classic Agent nodes choose only among tools configured at build time. The beta new Agent can also install and run command-line tools inside its sandbox when a task needs one, though anything installed during a published run is temporary.', + shortValue: 'Pre-wired tools; sandboxed agents can install CLIs', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/build/new-agent/build', + label: 'Dify Docs: Build an Agent (Tools)', + asOf: '2026-11-19', + }, + ], + }, + modelFallback: { + value: + 'Partial: LLM nodes retry failed calls with configurable attempts, intervals, and backoff, and a fail branch can route to a node using another model. Load balancing (Professional and Team) rotates API keys for the same model and rests a rate-limited key for a minute; no automatic switch to a different provider is documented.', + shortValue: 'Retries and key rotation; manual fallback via fail branch', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/nodes/llm', + label: 'Dify Docs: LLM node (Error Handling)', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/cloud/use-dify/workspace/model-providers', + label: 'Dify Docs: Model Providers (Load Balancing)', + asOf: '2026-11-19', + }, + ], + }, + agentSkills: { + value: + 'Yes: a workspace Skills library holds up to 500 skills in the open SKILL.md format, each with instructions plus supporting scripts and files, with draft, publish, and version history. Publishing an update propagates to every agent that uses the skill.', + shortValue: 'Workspace library of SKILL.md skills, up to 500', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/self-host/use-dify/build/skills', + label: 'Dify Docs: Skills', + asOf: '2026-11-19', + }, + ], + }, + nativeChatDeployment: { + value: + 'Yes: every published app gets a public web app link, and chat apps can be embedded on any website as a chat widget or iframe. Paid plans can restrict access with IP access control.', + shortValue: 'Yes, public web app plus embeddable chat widget', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/publish/README', + label: 'Dify Docs: Publish overview', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/cloud/use-dify/publish/webapp/embedding-in-websites', + label: 'Dify Docs: Embed Your Web App', + asOf: '2026-11-19', + }, + ], + }, + kbChunkVisibility: { + value: + "Yes: the Retrieval Testing page shows which chunks a query returns and logs retrievals from linked apps, and chunks can be added, edited, or disabled. In an agent's build chat, each knowledge lookup shows the passages it returned.", + shortValue: 'Yes, retrieval testing shows returned chunks', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/knowledge/test-retrieval', + label: 'Dify Docs: Test Knowledge Retrieval', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/cloud/use-dify/build/new-agent/build', + label: 'Dify Docs: Build an Agent (Check What the Agent Found)', + asOf: '2026-11-19', + }, + ], + }, + parallelExecution: { + value: + 'Yes: nodes can run in parallel branches whose results converge on a downstream node, and Iteration nodes can process up to 10 items at once in parallel mode. Self-hosted deployments default to 10 parallel branches per workflow.', + shortValue: 'Yes, parallel branches and parallel iteration', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/build/orchestrate-node', + label: 'Dify Docs: Orchestration Logic', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/cloud/use-dify/nodes/iteration', + label: 'Dify Docs: Iteration node', + asOf: '2026-11-19', + }, + ], + }, + a2aProtocol: { + value: + 'No: Dify core does not ship Agent2Agent (A2A) protocol support, and the GitHub feature request was closed as not planned in March 2026. Community Marketplace plugins offer an A2A server and client.', + shortValue: 'No native A2A; community plugins only', + confidence: 'estimated', + sources: [ + { + url: 'https://github.com/langgenius/dify/issues/19352', + label: 'GitHub langgenius/dify Issue #19352: Support for the A2A protocol', + asOf: '2026-11-19', + }, + ], + }, + loopIteration: { + value: + 'Yes: an Iteration node runs a sub-flow over each item in an array, sequentially or in parallel, with per-item error handling, and a Loop node repeats steps with persistent loop variables until a condition, a maximum count (100 by default on self-hosted deployments), or an Exit Loop node.', + shortValue: 'Yes, Iteration (for-each) and Loop (while) nodes', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/nodes/iteration', + label: 'Dify Docs: Iteration node', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/cloud/use-dify/nodes/loop', + label: 'Dify Docs: Loop node', + asOf: '2026-11-19', + }, + ], + }, + }, + integrations: { + integrationCount: { + value: + 'More than 1,000 plugins in Dify Marketplace, including roughly 945 tool plugins, 192 model provider plugins, and smaller sets of extensions and agent strategies, from Dify, partners, and community developers. Any OpenAPI service or HTTP MCP server can also be added as a tool.', + detail: 'Counts read from the Marketplace search on 2026-11-19 and change frequently.', + shortValue: '1,000+ Marketplace plugins, plus OpenAPI and MCP', + confidence: 'estimated', + sources: [ + { + url: 'https://marketplace.dify.ai', + label: 'Dify Marketplace', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/cloud/use-dify/workspace/tools', + label: 'Dify Docs: Dify Tools', + asOf: '2026-11-19', + }, + ], + }, + triggerTypes: { + value: + 'Workflow apps can start from a schedule trigger, a webhook trigger, or an integration trigger that subscribes to events from a plugin such as GitHub or Slack, as well as from the web app or API. Sandbox allows 2 triggers per workflow and 3,000 trigger events, Professional 20,000 events a month, and Team unlimited.', + shortValue: 'Schedule, webhook, integration events, API, web app', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/nodes/trigger/overview', + label: 'Dify Docs: Trigger', + asOf: '2026-11-19', + }, + { + url: 'https://dify.ai/pricing', + label: 'Dify Pricing (trigger events)', + asOf: '2026-11-19', + }, + ], + }, + customCodeSteps: { + value: + 'Yes: Code nodes run Python or JavaScript in a sandbox that blocks file system access and system commands, with up to 10 automatic retries and a fail branch, and HTTP Request and Template nodes cover API calls and Jinja2 formatting.', + shortValue: 'Yes, Python and JavaScript Code nodes', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/nodes/code', + label: 'Dify Docs: Code node', + asOf: '2026-11-19', + }, + ], + }, + codeSandboxRuntime: { + value: + 'Partial: Code nodes can import only a standard set of preinstalled packages, and imports outside it fail at run time. Self-hosted operators run the dify-sandbox service themselves and control its image and limits, and the beta new Agent can install programs inside its own sandbox.', + detail: + 'The self-hosted sandbox defaults to a 15-second limit per code run and routes outbound HTTP through an SSRF proxy.', + shortValue: 'Fixed package set; self-hosters control the sandbox', + confidence: 'estimated', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/nodes/code', + label: 'Dify Docs: Code node (Dependencies Support)', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/self-host/deploy/configuration/environments', + label: 'Dify Docs: Environment Variables (Sandbox Service)', + asOf: '2026-11-19', + }, + ], + }, + apiPublishing: { + value: + "Yes: every published app has a Service API endpoint, in blocking or streaming mode, plus endpoints to run a specific version, stop a run, resume a run's event stream, and list run logs. Dify publishes an OpenAPI spec for the Service API.", + shortValue: 'Yes, Service API with OpenAPI spec', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/api-reference/guides/workflow', + label: 'Dify Docs: Workflow App API', + asOf: '2026-11-19', + }, + ], + }, + extensibilitySdk: { + value: + 'Yes: a plugin SDK and CLI build tool, model, agent strategy, data source, trigger, and endpoint plugins, which can stay private or be published to Dify Marketplace. The repository also includes Node.js and PHP client SDKs, and the difyctl CLI manages apps from a terminal.', + shortValue: 'Plugin SDK, Marketplace, Node.js/PHP SDKs, CLI', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/develop-plugin/getting-started/getting-started-dify-plugin', + label: 'Dify Docs: Dify Plugin', + asOf: '2026-11-19', + }, + { + url: 'https://github.com/langgenius/dify/tree/main/sdks', + label: 'GitHub: langgenius/dify sdks directory', + asOf: '2026-11-19', + }, + ], + }, + mcpPublishing: { + value: + "Yes: each app's Access Point tab has an MCP Server toggle that generates a unique server URL, so clients such as Claude Desktop and Cursor can call the app as a tool. The URL contains credentials and can be regenerated.", + shortValue: 'Yes, any app can be exposed as an MCP server', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/publish/publish-mcp', + label: 'Dify Docs: MCP Server', + asOf: '2026-11-19', + }, + ], + }, + }, + pricing: { + pricingModel: { + value: + 'Per-workspace subscription on Dify Cloud with monthly message credits for built-in models, plus app, member, knowledge, and trigger quotas per plan. The self-hosted Community Edition is free under the Dify Open Source License, and Enterprise is custom-priced.', + shortValue: 'Per-workspace plans with message credits', + confidence: 'verified', + sources: [ + { + url: 'https://dify.ai/pricing', + label: 'Dify Pricing', + asOf: '2026-11-19', + }, + ], + }, + entryPaidPlan: { + value: + 'Professional: $59 per workspace per month, or $590 per year, with 3 team members, 50 apps, 5,000 message credits a month, 500 knowledge documents, and 5 GB of knowledge storage. Team is $159 per workspace per month ($1,590 per year) with 50 members.', + shortValue: 'Professional at $59/workspace/month', + confidence: 'verified', + sources: [ + { + url: 'https://dify.ai/pricing', + label: 'Dify Pricing', + asOf: '2026-11-19', + }, + { + url: 'https://dify.ai/pricing/dify-cloud', + label: 'Dify Cloud plan comparison', + asOf: '2026-11-19', + }, + ], + }, + freeTier: { + value: + 'Yes: the Sandbox plan is free with 200 message credits, 1 member, 5 apps, 50 knowledge documents, 50 MB of knowledge storage, 3,000 trigger events, and 30 days of log history. The self-hosted Community Edition is also free.', + shortValue: 'Sandbox plan: 200 credits, 5 apps, 1 member', + confidence: 'verified', + sources: [ + { + url: 'https://dify.ai/pricing', + label: 'Dify Pricing', + asOf: '2026-11-19', + }, + ], + }, + byok: { + value: + 'Yes: owners and admins add their own provider API keys, billed by the provider, alongside or instead of message credits, and can set credits to be spent first. Some models can only be used with your own key, and load balancing across several keys requires Professional or Team.', + shortValue: 'Yes, provider keys alongside message credits', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/workspace/model-providers', + label: 'Dify Docs: Model Providers', + asOf: '2026-11-19', + }, + ], + }, + }, + security: { + soc2: { + value: + "Yes: Dify's site footer shows a SOC 2 badge and its homepage lists SOC 2 Type II for Dify Enterprise. The Dify Cloud plan comparison offers a SOC 2 Type II report on the Team plan, though the Vanta-hosted Trust Center's description still refers to a SOC 2 Type 1 report available to paid customers by email.", + shortValue: 'SOC 2 Type II report (Team and Enterprise)', + confidence: 'verified', + sources: [ + { + url: 'https://dify.ai', + label: 'Dify homepage (Dify Enterprise: SOC 2 Type II + ISO 27001)', + asOf: '2026-11-19', + }, + { + url: 'https://dify.ai/pricing/dify-cloud', + label: 'Dify Cloud plan comparison (SOC Type II Report)', + asOf: '2026-11-19', + }, + { + url: 'https://security.dify.ai', + label: 'Dify Trust Center', + asOf: '2026-11-19', + }, + ], + }, + dataResidency: { + value: + 'Via self-hosting or Enterprise: Dify Enterprise deploys inside your own VPC, region, on premises, or air-gapped. Dify Cloud stores data in its managed cloud region, and its privacy policy says website servers are in the United States; no customer-selectable Cloud region is documented.', + shortValue: 'Self-host or Enterprise VPC; no Cloud region choice', + confidence: 'verified', + sources: [ + { + url: 'https://dify.ai/dify-enterprise', + label: 'Dify Enterprise page (Data residency)', + asOf: '2026-11-19', + }, + { + url: 'https://dify.ai/pricing/dify-cloud', + label: 'Dify Cloud FAQ (Where is my data stored?)', + asOf: '2026-11-19', + }, + { + url: 'https://dify.ai/legal/privacy-policy', + label: 'Dify Privacy Policy (Data Transfers)', + asOf: '2026-11-19', + }, + ], + }, + rbac: { + value: + 'Yes: workspaces have four built-in roles (Owner, Admin, Editor, Normal), and each knowledge base has its own access settings. Dify Enterprise adds custom workspace roles and permission templates, and Dify describes its Enterprise RBAC as fine-grained down to the workflow level.', + detail: + 'Dify Cloud has role management on Professional and Team; the plan comparison does not list it for the free Community Edition.', + shortValue: 'Four workspace roles; custom roles on Enterprise', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/workspace/team-members-management', + label: 'Dify Docs: Manage Members', + asOf: '2026-11-19', + }, + { + url: 'https://enterprise-docs.dify.ai/en/3.13.x/administer/admin-console-rbac-help', + label: 'Dify Enterprise Docs: Configure Permission Templates', + asOf: '2026-11-19', + }, + ], + }, + auditLogging: { + value: + 'Enterprise only: the Admin Dashboard audit log records operator, IP address, operation, and resource for actions such as logins, app publishing, DSL import and export, and permission changes, with 90 days of online search and CSV download for older logs. Dify describes the log as tamper-evident and streamable to a SIEM.', + detail: + 'Professional and Team admins can view operation logs of individual workspace members.', + shortValue: 'Enterprise audit log with CSV export', + confidence: 'verified', + sources: [ + { + url: 'https://enterprise-docs.dify.ai/en/3.13.x/administer/audit-log', + label: 'Dify Enterprise Docs: Audit Log', + asOf: '2026-11-19', + }, + { + url: 'https://dify.ai/dify-enterprise', + label: 'Dify Enterprise page (Audit & observability)', + asOf: '2026-11-19', + }, + ], + }, + additionalCompliance: { + value: + 'Dify lists ISO 27001 certification and GDPR alongside SOC 2 in its site footer and homepage, publishes a Data Protection Agreement, and offers penetration test reports to paid customers through its Trust Center. No HIPAA, PCI, or FedRAMP attestation is published.', + shortValue: 'ISO 27001, GDPR, and SOC 2; no HIPAA published', + confidence: 'verified', + sources: [ + { + url: 'https://dify.ai', + label: 'Dify homepage (compliance badges)', + asOf: '2026-11-19', + }, + { + url: 'https://security.dify.ai', + label: 'Dify Trust Center', + asOf: '2026-11-19', + }, + ], + }, + modelAndToolGovernance: { + value: + 'Partial: owners and admins decide who may install and debug plugins, and Dify Enterprise can limit installs to official or official-plus-partner plugins, Marketplace sources only, or signed packages, and push a plugin version to every workspace. No per-role allow or deny list for specific models or tools is documented.', + shortValue: 'Plugin install policies; no per-role model allow-lists', + confidence: 'estimated', + sources: [ + { + url: 'https://enterprise-docs.dify.ai/en/3.13.x/administer/plugins-management', + label: 'Dify Enterprise Docs: Plugin installation strategy', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/cloud/use-dify/workspace/plugins', + label: 'Dify Docs: Integrations (Permissions and Updates)', + asOf: '2026-11-19', + }, + ], + }, + credentialGovernance: { + value: + 'Yes, on Enterprise: administrators create global model and tool plugin credentials, assign them to specific workspaces, and set credential policies that allow or forbid workspaces from creating their own credentials for a plugin. In Community and Cloud workspaces, only owners and admins manage model provider keys.', + shortValue: 'Enterprise global credentials and credential policies', + confidence: 'verified', + sources: [ + { + url: 'https://enterprise-docs.dify.ai/en/3.13.x/administer/credential-management', + label: 'Dify Enterprise Docs: Credential Management', + asOf: '2026-11-19', + }, + ], + }, + whiteLabeling: { + value: + 'Partial: web app branding with a custom logo and UI is on the Team plan and Enterprise, and Enterprise branding settings replace the application title, login page logo, and workspace logo. On self-hosted Community deployments, the license forbids removing or modifying the Dify logo in the console or applications without a commercial license.', + shortValue: 'Team web app branding; full rebrand on Enterprise', + confidence: 'verified', + sources: [ + { + url: 'https://enterprise-docs.dify.ai/en/3.13.x/administer/branding', + label: 'Dify Enterprise Docs: Branding Settings', + asOf: '2026-11-19', + }, + { + url: 'https://dify.ai/pricing/dify-cloud', + label: 'Dify Cloud plan comparison (WebApp Branding Customization)', + asOf: '2026-11-19', + }, + { + url: 'https://github.com/langgenius/dify/blob/main/LICENSE', + label: 'langgenius/dify LICENSE (GitHub)', + asOf: '2026-11-19', + }, + ], + }, + dataRetention: { + value: + 'Partial: Dify Cloud keeps logs for 30 days on Sandbox; on Professional and Team, conversation logs last for the subscription and workflow run logs older than about three months move to downloadable monthly CSV archives. Self-hosted deployments can turn on automatic workflow log cleanup with a configurable retention period (30 days by default).', + shortValue: 'Fixed Cloud retention; configurable when self-hosted', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/monitor/logs', + label: 'Dify Docs: Logs (Log Retention)', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/self-host/deploy/configuration/environments', + label: 'Dify Docs: Environment Variables (Workflow Log Cleanup)', + asOf: '2026-11-19', + }, + ], + }, + piiRedaction: { + value: + 'Partial: Dify Enterprise advertises PII-redacted prompt history, and Enterprise telemetry export replaces input and output content with reference IDs by default. Cloud and Community apps rely on content moderation (OpenAI Moderation, keywords, or a custom endpoint) rather than a documented PII detection feature.', + shortValue: 'Enterprise PII-redacted history; moderation elsewhere', + confidence: 'estimated', + sources: [ + { + url: 'https://dify.ai/dify-enterprise', + label: 'Dify Enterprise page (Audit & observability)', + asOf: '2026-11-19', + }, + { + url: 'https://enterprise-docs.dify.ai/en/3.13.x/administer/data-push', + label: 'Dify Enterprise Docs: Push Telemetry (Control Input/Output Content)', + asOf: '2026-11-19', + }, + ], + }, + sso: { + value: + 'Enterprise only: SSO through SAML, OIDC, or OAuth2 for workspace members, internal web app users, and external web app users, with SCIM 2.0 member sync from providers such as Okta and Microsoft Entra ID. SAML and OIDC SSO are not offered on Dify Cloud or the Community Edition.', + shortValue: 'SAML/OIDC/OAuth2 SSO and SCIM on Enterprise', + confidence: 'verified', + sources: [ + { + url: 'https://enterprise-docs.dify.ai/en/3.13.x/administer/sso/introduction', + label: 'Dify Enterprise Docs: SSO Authentication Introduction', + asOf: '2026-11-19', + }, + { + url: 'https://enterprise-docs.dify.ai/en/3.13.x/administer/members/synchronization/sync-team-members', + label: 'Dify Enterprise Docs: Sync Members (SCIM)', + asOf: '2026-11-19', + }, + { + url: 'https://dify.ai/pricing/dify-enterprise', + label: 'Dify Pricing: Self-Hosted and Enterprise plan comparison', + asOf: '2026-11-19', + }, + ], + }, + sessionPolicy: { + value: + 'Partial: self-hosted operators set access-token lifetime (60 minutes by default) and how long a user stays signed in without visiting (30 days by default) through environment variables. Dify lists session policies among Enterprise identity features, but no admin console setting for them is documented publicly.', + shortValue: 'Self-hosted token env variables; Enterprise claims policies', + confidence: 'estimated', + sources: [ + { + url: 'https://docs.dify.ai/en/self-host/deploy/configuration/environments', + label: 'Dify Docs: Environment Variables (Token & Request Limits)', + asOf: '2026-11-19', + }, + { + url: 'https://dify.ai/dify-enterprise', + label: 'Dify Enterprise page (Identity & access)', + asOf: '2026-11-19', + }, + ], + }, + thirdPartyVetting: { + value: + 'Partial: Dify says Marketplace plugins from Dify, partners, and community developers are reviewed, tested, and signed before publication, then maintained by their publishers. By default, workspaces can also install plugins from any public GitHub repository or a local package; only Dify Enterprise can restrict installs to official, partner, Marketplace, or signed plugins.', + detail: + "Sim's integrations are all first-party blocks reviewed through Sim's own pull-request process, with no third-party marketplace of installable code.", + shortValue: 'Reviewed Marketplace, but GitHub and local installs allowed', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/workspace/plugins', + label: 'Dify Docs: Integrations (Install Integrations)', + asOf: '2026-11-19', + }, + { + url: 'https://enterprise-docs.dify.ai/en/3.13.x/administer/plugins-management', + label: 'Dify Enterprise Docs: Plugin installation strategy', + asOf: '2026-11-19', + }, + ], + }, + }, + observability: { + tracingDepth: { + value: + 'Yes: each workflow run shows inputs, outputs, status, token totals, and a node-by-node trace, including drill-down into nested workflow tools, and each chat answer shows its token spend, latency, and raw prompt. Traces export to LangSmith, Langfuse, Opik, W&B Weave, Arize, Phoenix, Alibaba Cloud Monitor, or, since 1.20.0, any OpenTelemetry collector.', + shortValue: 'Node-level run traces, exportable to LLM tracing tools', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/monitor/logs', + label: 'Dify Docs: Logs', + asOf: '2026-11-19', + }, + { + url: 'https://github.com/langgenius/dify/releases/tag/1.20.0', + label: 'Dify v1.20.0 release notes (GitHub)', + asOf: '2026-11-19', + }, + ], + }, + durabilityModel: { + value: + 'Partial: LLM, HTTP Request, and Code nodes retry failed attempts automatically (HTTP and Code up to 10 times), Human Input pauses persist until a response or timeout, and a past run can be re-run in the editor with its original inputs through Test With Params. No checkpoint-and-resume of a failed run from the failing step is documented.', + shortValue: 'Node retries, persisted pauses, re-run with inputs', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/nodes/code', + label: 'Dify Docs: Code node (Error Handling and Retries)', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/cloud/use-dify/monitor/logs', + label: 'Dify Docs: Logs (Inspect a Run)', + asOf: '2026-11-19', + }, + ], + }, + failureAlerting: { + value: + "Unknown: Dify's documentation describes no built-in notification when a run fails or crosses a cost or latency threshold. Builders can wire a fail branch to send a notification, and Enterprise customers can set alerts on the telemetry they export to their own collector.", + shortValue: 'No built-in failure alerts documented', + confidence: 'estimated', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/build/predefined-error-handling-logic', + label: 'Dify Docs: Handle Errors', + asOf: '2026-11-19', + }, + ], + }, + dataDrains: { + value: + 'Yes: apps continuously export traces to LLM observability tools or any OpenTelemetry collector, and Dify Enterprise Data Push streams metrics and traces over OTLP to your own collector. Workflow run logs older than about three months are also available as monthly CSV archives on paid Cloud plans.', + shortValue: 'Trace export to OTel and LLM tools; Enterprise OTLP push', + confidence: 'verified', + sources: [ + { + url: 'https://enterprise-docs.dify.ai/en/3.13.x/administer/data-push', + label: 'Dify Enterprise Docs: Push Telemetry to Your Observability Stack', + asOf: '2026-11-19', + }, + { + url: 'https://github.com/langgenius/dify/releases/tag/1.20.0', + label: 'Dify v1.20.0 release notes (GitHub)', + asOf: '2026-11-19', + }, + ], + }, + asyncExecution: { + value: + 'Partial: webhook triggers return 200 OK once the workflow is triggered, and an API run returns a workflow run ID whose status and outputs can be fetched later or whose event stream can be resumed. The Run Workflow API itself offers streaming or blocking modes rather than a fire-and-forget mode.', + shortValue: 'Webhook triggers plus run-status lookup and stream resume', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/api-reference/guides/workflow', + label: 'Dify Docs: Workflow App API', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/cloud/use-dify/nodes/trigger/webhook-trigger', + label: 'Dify Docs: Webhook Trigger', + asOf: '2026-11-19', + }, + ], + }, + executionLimits: { + value: + 'Self-hosted defaults: 1 hour per workflow run, 500 node executions per run, 10 parallel branches, 100 loop iterations, and 15 seconds per code run, all configurable; a single execution path holds up to 50 nodes. New Agent runs stop after 1 hour or 500 model requests. Dify Cloud does not publish per-run time or concurrency limits beyond plan quotas and queue priority.', + shortValue: '1 hour and 500 steps per run by default (self-hosted)', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/self-host/deploy/configuration/environments', + label: 'Dify Docs: Environment Variables (Workflow Runtime)', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/cloud/use-dify/build/new-agent/build', + label: 'Dify Docs: Build an Agent (Run Limits)', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/cloud/use-dify/build/orchestrate-node', + label: 'Dify Docs: Orchestration Logic', + asOf: '2026-11-19', + }, + ], + }, + partialFailureHandling: { + value: + 'Yes: LLM, HTTP, Code, and Tool nodes can substitute a default value or follow a fail branch with error type and message variables, and Iteration nodes can skip or drop failed items. Loop nodes always stop when a child node fails.', + shortValue: 'Yes, fail branches and default values per node', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/build/predefined-error-handling-logic', + label: 'Dify Docs: Handle Errors', + asOf: '2026-11-19', + }, + ], + }, + unattendedExecution: { + value: + 'Yes: schedule, webhook, and integration triggers run workflows on the Dify server, on Dify Cloud or a self-hosted deployment, with no dependency on a client device staying open. A published trigger can be paused from the Access Point tab.', + shortValue: 'Yes, runs server-side on Cloud or self-hosted', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/cloud/use-dify/nodes/trigger/overview', + label: 'Dify Docs: Trigger', + asOf: '2026-11-19', + }, + ], + }, + }, + support: { + supportChannels: { + value: + 'Sandbox and Community users get documentation, GitHub Discussions and Issues, and Discord; Professional and Team add priority email support. Dify Enterprise includes standard technical support with paid one-on-one support, and Dify lists a named customer success manager, a Slack Connect channel, and 24/7 support for Enterprise.', + shortValue: 'Community, priority email, Enterprise CSM and 24/7', + confidence: 'verified', + sources: [ + { + url: 'https://dify.ai/pricing/dify-cloud', + label: 'Dify Cloud plan comparison (Support)', + asOf: '2026-11-19', + }, + { + url: 'https://dify.ai/dify-enterprise', + label: 'Dify Enterprise page (Enterprise Support Agreements)', + asOf: '2026-11-19', + }, + ], + }, + sla: { + value: + "Enterprise only: Dify's Enterprise page lists a 1-hour response on Sev-1 issues, and the Enterprise plan comparison lists negotiated SLAs delivered by Dify partners. No SLA is published for Dify Cloud plans.", + shortValue: 'Enterprise: 1-hour Sev-1 response, negotiated SLAs', + confidence: 'verified', + sources: [ + { + url: 'https://dify.ai/dify-enterprise', + label: 'Dify Enterprise page (Enterprise Support Agreements)', + asOf: '2026-11-19', + }, + { + url: 'https://dify.ai/pricing/dify-enterprise', + label: 'Dify Pricing: Self-Hosted and Enterprise plan comparison', + asOf: '2026-11-19', + }, + ], + }, + community: { + value: + "Dify's GitHub repository has about 161,500 stars and 25,500 forks, with active GitHub Discussions, a Discord server, and a subreddit. Dify says it runs on more than 1.4 million machines.", + shortValue: 'About 161,500 GitHub stars, 25,500 forks', + confidence: 'verified', + sources: [ + { + url: 'https://github.com/langgenius/dify', + label: 'GitHub: langgenius/dify', + asOf: '2026-11-19', + }, + { + url: 'https://dify.ai/about-us', + label: 'Dify About Us', + asOf: '2026-11-19', + }, + ], + }, + companyMaturity: { + value: + 'Dify is built by LangGenius, Inc.; its public repository dates to April 2023. LangGenius reports a $30M Series Pre-A round in 2026 at a $180M valuation led by HSG, more than 280 enterprise customers, and a Niche Player placement in the 2026 Gartner Magic Quadrant for AI Agent Management Platforms.', + shortValue: 'Since 2023; $30M Pre-A in 2026, 280+ enterprise customers', + confidence: 'verified', + sources: [ + { + url: 'https://dify.ai/about-us', + label: 'Dify About Us', + asOf: '2026-11-19', + }, + { + url: 'https://dify.ai/pricing', + label: 'Dify Pricing (Gartner Magic Quadrant banner)', + asOf: '2026-11-19', + }, + { + url: 'https://github.com/langgenius/dify', + label: 'GitHub: langgenius/dify', + asOf: '2026-11-19', + }, + ], + }, + academy: { + value: + 'Partial: the docs include a 10-lesson Workflow 101 course and step-by-step tutorials such as a knowledge base support bot and an image generation app. No certification program is documented, and Dify for Education gives verified students and teachers a free year of the Professional plan.', + shortValue: 'Workflow 101 course and tutorials; no certification', + confidence: 'verified', + sources: [ + { + url: 'https://docs.dify.ai/en/learn/tutorials/workflow-101/lesson-01', + label: 'Dify Docs: Workflow 101, Lesson 1', + asOf: '2026-11-19', + }, + { + url: 'https://docs.dify.ai/en/cloud/use-dify/workspace/subscription-management', + label: 'Dify Docs: Billing (Dify for Education)', + asOf: '2026-11-19', + }, + ], + }, + }, + }, +} diff --git a/apps/sim/lib/compare/data/competitors/relevance-ai.ts b/apps/sim/lib/compare/data/competitors/relevance-ai.ts new file mode 100644 index 00000000000..83433b0defc --- /dev/null +++ b/apps/sim/lib/compare/data/competitors/relevance-ai.ts @@ -0,0 +1,1416 @@ +import { RelevanceAIIcon } from '@/components/icons' +import type { CompetitorProfile } from '@/lib/compare/data/types' + +/** Researched and cross-verified against live vendor sources on 2026-11-19. */ +export const relevanceAiProfile: CompetitorProfile = { + id: 'relevance-ai', + name: 'Relevance AI', + website: 'https://relevanceai.com', + brand: { + icon: RelevanceAIIcon, + selfFramed: false, + colors: ['#4F46E5', '#818CF8', '#C7D2FE', '#0D162F'], + source: 'Relevance AI official logo SVG (relevanceai.com/images/redesign/relevance-logo.svg)', + asOf: '2026-11-19', + }, + oneLiner: + 'Relevance AI is a hosted, proprietary platform for building and running teams of AI agents (an "AI Workforce"), with a no-code agent and tool builder, a visual multi-agent Workforce builder, a built-in building agent called Invent, and packaged agents aimed at sales, marketing, and customer success teams.', + leadAnswer: [ + 'Relevance AI is a hosted AI agent platform built around the "AI Workforce": specialist agents that hand work to each other in a visual multi-agent builder, with approvals, evals, and packaged agents for go-to-market teams. Sim is the open-source AI workspace where teams build, deploy, and manage AI agents visually, conversationally, or with code. Choose Relevance AI when a sales, marketing, or customer success team wants ready-made GTM agents, phone and meeting agents, and a managed service it never has to run. Choose Sim when you need to self-host, read the source, edit the same workflow live with teammates, or promote agents across dev, QA, and production workspaces.', + ], + betterThanAnswer: [ + 'Sim is the stronger fit when the team building agents needs control over where they run and how they are promoted: its core is Apache 2.0 and self-hostable, and its Enterprise plan adds workspace forking with diff and promote. Relevance AI is the stronger fit for go-to-market teams that want a managed multi-agent service with packaged sales agents, phone and meeting agents, and pre-publish evals, and that accept hosted-only deployment and Enterprise-gated governance. Choose on that axis: an open, self-hostable workspace, or a managed agent service focused on GTM work.', + ], + sectionIntros: { + platform: [ + 'Sim offers a visual workflow builder, Chat, and code, with supported self-hosting, live multiplayer editing, and workspace promotion. Relevance AI provides a hosted no-code agent and tool builder plus a visual Workforce builder for multi-agent teams, with draft-and-publish version history.', + ], + aiCapabilities: [ + 'Both products support many model providers, knowledge bases, MCP, and human approvals. Relevance AI centers its model on teams of agents with per-connection approval modes and Enterprise-tier evals; Sim exposes evaluation, approvals, loops, and parallel branches as blocks in the workflow itself.', + ], + integrations: [ + 'Relevance AI cites 2,000+ integrations, most of them added through a Pipedream connection, plus custom API steps, an MCP server, and a JavaScript SDK. Sim connects its own integration catalog, publishes deployed workflows as MCP tools, and offers a REST API, OpenAPI spec, and CLI.', + ], + pricing: [ + 'Sim charges for usage through credits, with bring-your-own-key calls skipping model charges. Relevance AI meters two things on every paid plan: Actions (one per Tool run) and Vendor Credits for model and tool costs, with bring-your-own keys bypassing Vendor Credits but not Actions.', + ], + security: [ + 'Both products state SOC 2 compliance. Relevance AI offers a choice of US, EU, or Australian data region at signup, while SSO, RBAC, audit logs, model controls, and custom retention are Enterprise-only; Sim adds self-hosting as its main data-control option.', + ], + observability: [ + 'Relevance AI documents task-status alerts, rerun from the failing step, OpenTelemetry export to S3 on Enterprise, and published run-time ceilings. Sim documents block-level traces, retries, failure alerts, data export, asynchronous runs, and execution limits.', + ], + support: [ + 'Relevance AI offers a community forum, in-app and email support, a free Academy, and a dedicated Slack channel for Enterprise. Sim offers an open-source community and an Enterprise dedicated-support option.', + ], + }, + faqs: [ + { + question: 'Can I self-host Relevance AI, or do I need a platform like Sim?', + answer: + "Relevance AI cannot be self-hosted: it runs as a multi-tenant hosted service, its security page says single-tenant options are still in the works, and its code is not published. Sim's core is Apache 2.0 licensed and can be self-hosted with Docker Compose or Kubernetes (Helm), so your team runs the whole platform. Relevance AI's answer to data location is a region choice at signup (US, EU, or Australia), which cannot be changed later.", + }, + { + question: 'How does Relevance AI pricing compare to Sim?', + answer: + "Relevance AI's Pro plan is $19 per month billed annually or $29 billed monthly, with 2,500 Actions and 10,000 Vendor Credits a month, and Team is $234 per month billed annually or $349 monthly; the Free plan has been retired. Every Tool run counts as an Action, even if it fails, and extra Actions cost $80 per 1,000. Sim's Pro plan is $25 per user per month with credit-based usage, and bring-your-own-key calls skip Sim's model usage charge.", + }, + { + question: 'Sim vs Relevance AI for multi-agent systems: which is better?', + answer: + "It depends on how much you want the platform to decide. Relevance AI's Workforce builder connects agents, tools, and conditions, and each connection can auto-run, require approval, or let the agent decide whether to hand off, with Invent available to build the team from a process description. Sim builds multi-agent systems as workflows with Agent blocks, loops, parallel branches, and approval blocks, built visually, conversationally, or with code, and deployable as a REST API, chat interface, or MCP server.", + }, + ], + standoutFeatures: [ + { + title: 'A visual multi-agent Workforce builder with per-connection approval modes', + description: + 'A Workforce connects specialist agents, tools, triggers, and conditions in a visual builder. Each connection is either an AI handoff the agent chooses or a mandatory next step, and each agent-to-tool or agent-to-agent edge can be set to Auto run (optionally capped by "Max auto runs"), Approval required, or Let agent decide, with pending approvals reviewed in the Workforce task view.', + shortDescription: + 'Visual multi-agent teams where each connection auto-runs, needs approval, or lets the agent decide.', + source: { + url: 'https://relevanceai.com/docs/build/workforces/workforce-features/approvals-and-escalations', + label: 'Approvals and Escalations | Relevance AI Docs', + asOf: '2026-11-19', + }, + }, + { + title: 'Invent, a built-in agent that builds and maintains the agent team', + description: + 'Invent takes a business process described in plain language and creates the Agents, Tools, Triggers, Knowledge tables, Workforces, and Evals to run it, then reviews live performance, diagnoses quality problems, and proposes tested fixes. It can also run scheduled routines and auto-diagnose failed triggers, with per-tool approval settings (Auto, Ask, Always ask) that default destructive actions such as delete and publish to Always ask.', + shortDescription: + 'Invent builds agents, tools, triggers, and evals from a process description, then monitors them.', + source: { + url: 'https://relevanceai.com/docs/build/invent/invent', + label: 'Build with Invent | Relevance AI Docs', + asOf: '2026-11-19', + }, + }, + { + title: 'Evals that simulate conversations and can block publishing', + description: + 'Evals runs Test sets that simulate users against an Agent or Workforce, scores each conversation with reusable Checks, and can require a minimum pass rate before a new version publishes. Monitor dashboards then score a sampled share of live conversations against the same Checks. Agent Evaluations are listed as an Enterprise-plan feature.', + shortDescription: + 'Simulated-conversation evals with a pass-rate gate on publishing and live sampling (Enterprise).', + source: { + url: 'https://relevanceai.com/docs/build/agents/evals/running-evaluations', + label: 'Running Evals | Relevance AI Docs', + asOf: '2026-11-19', + }, + }, + { + title: 'Phone and meeting agents', + description: + 'Phone Agents make outbound calls, speak with the person on the line, and run tools during the call, with a Relevance-provided or your own phone number. Meeting agents join calls to take notes and respond to trigger words typed in the meeting chat. Calling and Meeting agent modes are included on the Team and Enterprise plans.', + shortDescription: + 'Voice agents that call out and act mid-call, plus meeting agents, on Team and Enterprise.', + source: { + url: 'https://relevanceai.com/docs/build/agents/build-your-agent/agent-modes/phone-calls', + label: 'Phone Agents | Relevance AI Docs', + asOf: '2026-11-19', + }, + }, + { + title: 'Build agents from Claude Code, Codex, or any MCP client', + description: + "Relevance AI's MCP server at mcp.relevanceai.com gives Claude Desktop, ChatGPT, Cursor, VS Code, and other MCP clients access to a project's agents, tools, and knowledge, and a Claude Code plugin and agent-skills repository let a coding agent create and edit Agents, Tools, Triggers, and Workforces in natural language. An official JavaScript/TypeScript SDK (@relevanceai/sdk) covers embedding agents in your own apps.", + shortDescription: + 'MCP server, Claude Code plugin, and a JS/TS SDK for building and running agents from code tools.', + source: { + url: 'https://relevanceai.com/docs/get-started/core-concepts/mcp-plugins', + label: 'MCP & Plugins | Relevance AI Docs', + asOf: '2026-11-19', + }, + }, + { + title: 'Choice of US, EU, or Australian data region', + description: + 'Data is stored in US (N. Virginia), EU (London), or AU (Sydney), chosen when the organization is created. The region cannot be changed afterwards without contacting support.', + shortDescription: 'Pick a US, EU, or Australian data region at signup.', + source: { + url: 'https://relevanceai.com/docs/admin/security', + label: 'Security overview | Relevance AI Docs', + asOf: '2026-11-19', + }, + }, + ], + limitations: [ + { + title: 'Hosted-only and proprietary, with no self-hosting option', + description: + "Relevance AI runs on a multi-tenant architecture with logical data separation, and its security page says single-tenant options are 'currently in the works'. No self-hosted or on-premises deployment is documented, and the platform's source code is not published.", + shortDescription: + 'Multi-tenant SaaS only; no self-hosting, and single-tenant is not yet offered.', + source: { + url: 'https://relevanceai.com/docs/admin/security', + label: 'Security overview | Relevance AI Docs', + asOf: '2026-11-19', + }, + }, + { + title: 'Governance and evals are Enterprise-only, and some are still rolling out', + description: + "SSO, RBAC, audit logs, Agent Evaluations, model controls, integration controls, and custom data retention are all listed as Enterprise-only. The RBAC, RBAC Groups, user-level authentication, and asset-level credential controls pages say these features are 'gradually being rolled out' to Enterprise customers.", + shortDescription: + 'SSO, RBAC, audit logs, and evals need Enterprise; RBAC is still rolling out.', + source: { + url: 'https://relevanceai.com/docs/get-started/pricing', + label: 'Pricing | Relevance AI Docs', + asOf: '2026-11-19', + }, + }, + { + title: 'Two usage meters, and every Tool run costs an Action even when it fails', + description: + 'Paid plans meter Actions (one per Tool run) and Vendor Credits (model and tool costs) separately. A failed Tool run still counts as one Action, plan Actions reset at each renewal, and extra Actions cost $80 per 1,000. Bringing your own model keys bypasses Vendor Credits but not Actions.', + shortDescription: + 'Actions and Vendor Credits are billed separately; failed Tool runs still use an Action.', + source: { + url: 'https://relevanceai.com/docs/admin/subscriptions/plans', + label: 'Plans and credits | Relevance AI Docs', + asOf: '2026-11-19', + }, + }, + { + title: 'No free plan, and the public pricing page lists only Enterprise', + description: + 'The Free plan is retired for new signups and existing accounts, so every organization needs a paid plan. On 2026-11-19 the relevanceai.com/pricing page showed only a sales-led Enterprise tier; Pro and Team prices appear on the pricing page in the documentation.', + shortDescription: 'Free plan retired; self-serve prices are only listed in the docs.', + source: { + url: 'https://relevanceai.com/docs/get-started/pricing', + label: 'Pricing | Relevance AI Docs', + asOf: '2026-11-19', + }, + }, + { + title: 'The API is supported only for triggering Agents and Tools', + description: + "Relevance AI's docs say the API is 'officially supported only for triggering Agents and Tools. All other usage is currently unsupported.' The agent API trigger does not return the agent's result in the response; the result has to be sent to a destination you configure, such as a webhook or Google Sheets.", + shortDescription: + 'API is supported only for triggering runs; results go to a configured destination.', + source: { + url: 'https://relevanceai.com/docs/build/agents/build-your-agent/agent-triggers/api-trigger', + label: 'API Triggers | Relevance AI Docs', + asOf: '2026-11-19', + }, + }, + ], + facts: { + platform: { + builderType: { + value: + 'No-code agent and tool builder (prompt, tools, triggers, knowledge), a visual Workforce builder that connects agents, tools, triggers, and conditions, plus Invent for natural-language building and an MCP server for coding agents', + detail: + 'Agents are configured through prompt, tools, triggers, and settings tabs; Tools are built from ordered steps with loops and branches; Workforces lay out multi-agent teams as connected nodes (triggers, agents, tools, conditions). Invent and the MCP server let a builder create the same assets in plain language.', + shortValue: 'No-code agent builder plus visual multi-agent Workforce builder', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/build/workforces/build-an-ai-workforce/introduction-to-nodes', + label: 'Introduction to Nodes | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/features', + label: 'Relevance AI Features', + asOf: '2026-11-19', + }, + ], + }, + learningCurve: { + value: + 'Low for a single agent or a cloned Marketplace agent; steeper for multi-agent Workforces with edge settings, Python tool steps, and Enterprise RBAC setup', + detail: + "Relevance AI positions the product for domain experts rather than engineers and offers a 5-minute quick start and a free Academy. Its own org-structure guide notes that environment-separated projects are 'more complex for non-technical users'.", + shortValue: 'Easy for single agents, steeper for Workforces and code steps', + confidence: 'estimated', + sources: [ + { + url: 'https://relevanceai.com/docs/get-started/quick-start-guide', + label: 'Quick Start | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/enterprise/org-structure-best-practices', + label: 'Organization Structure Best Practices | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + selfHostOption: { + value: + 'No: Relevance AI is offered only as a hosted, multi-tenant service; its security page says single-tenant options are still in the works', + detail: + 'Customer data is logically separated in a multi-tenant architecture, with a separate service and database for Enterprise customers that enable fine-grained access controls. No self-hosted or on-premises installation is documented.', + shortValue: 'No, hosted multi-tenant SaaS only', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/admin/security', + label: 'Security overview | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + deploymentOptions: { + value: + 'Hosted cloud in a US, EU, or Australian region chosen at signup; agents run through the web app, Relevance Chat (web, desktop, iOS, Android), Slack and Teams triggers, an embeddable chat widget, API, and SDK', + detail: + 'The data region (US N. Virginia, EU London, or AU Sydney) is fixed when the organization is created. Agents can also be shared as a public Chat UI link, an embeddable chat widget, or a Task UI iframe.', + shortValue: 'Hosted only, US/EU/AU region; chat, widget, API, SDK', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/admin/security', + label: 'Security overview | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/build/agents/share-your-agent', + label: 'Share Your Agent | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + templates: { + value: + 'Yes: a Marketplace of agents and tools built by Relevance AI staff and vetted community builders, some free and some paid, plus cloneable agent and Workforce templates', + detail: + 'Marketplace listings can be free or paid (creators set prices up to $1,000), and every submission is reviewed by the Relevance AI team. Agents and Workforces can also be shared as cloneable templates that run on the recipient’s own credits.', + shortValue: 'Marketplace of free and paid agents, plus cloneable templates', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/get-started/marketplace/introduction', + label: 'Marketplace | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/build/workforces/share-your-workforce', + label: 'Share your Workforce | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + license: { + value: 'Proprietary, closed-source SaaS', + detail: + "Relevance AI is operated by OnSearch Pty Ltd T/A Relevance AI. Its public GitHub organization hosts documentation, not the platform's source, and no open-source license for the product is published.", + shortValue: 'Proprietary', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/about-us', + label: 'About us | Relevance AI', + asOf: '2026-11-19', + }, + { + url: 'https://github.com/RelevanceAI/relevance-docs', + label: 'RelevanceAI/relevance-docs (GitHub)', + asOf: '2026-11-19', + }, + ], + }, + environmentPromotion: { + value: + 'No formal promotion: the docs suggest separate Development, Staging, and Production projects, with assets copied between them using "Copy to Project", and note you "may need to duplicate assets across environments"', + detail: + 'Within one project, edits go to a draft and only reach the live version on publish, which separates work-in-progress from production for a single asset. There is no documented diff-and-promote flow between whole projects.', + shortValue: 'Separate projects per environment, assets copied manually', + confidence: 'estimated', + sources: [ + { + url: 'https://relevanceai.com/docs/enterprise/org-structure-best-practices', + label: 'Organization Structure Best Practices | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/build/agents/share-your-agent', + label: 'Share Your Agent | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + versionControlDepth: { + value: + 'Every draft save and publish creates a version for Agents, Tools, and Workforces; any version can be restored into a draft and renamed, and the live version keeps running until you publish', + detail: + 'The version panel shows who saved each version and marks versions made by Invent or over MCP. Versions are kept indefinitely unless an Enterprise retention policy applies, and the active and draft versions are always protected. No side-by-side diff view is documented.', + shortValue: 'Auto-versioned drafts and publishes with restore', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/build/agents/build-your-agent/version-history', + label: 'Version history | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + realtimeCollaboration: { + value: + 'Not documented: projects are shared by members with roles, but no live co-editing of the same agent or Workforce with shared cursors is described', + detail: + "Relevance AI's collaboration page and docs cover shared projects, roles, sharing, and escalations rather than live concurrent editing.", + shortValue: 'Shared projects; live co-editing not documented', + confidence: 'estimated', + sources: [ + { + url: 'https://relevanceai.com/collaboration', + label: 'Collaboration | Relevance AI', + asOf: '2026-11-19', + }, + ], + }, + nativeFileStorage: { + value: + 'Not documented: files are uploaded into Knowledge tables or Chat conversations; no standalone file manager with folders and link sharing is described', + detail: + 'Uploads are capped at 100 MB, and Knowledge accepts CSV, PDF, Excel, JSON, and audio files.', + shortValue: 'No native file manager documented', + confidence: 'estimated', + sources: [ + { + url: 'https://relevanceai.com/docs/admin/system-limits', + label: 'System quotas | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/build/knowledge/create-knowledge', + label: 'Create and Organize Knowledge | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + dataTables: { + value: + 'Partial: Knowledge tables are native tables with columns you define and rows agents can insert, update, delete, and enrich with AI-generated columns; spreadsheet keyboard navigation is not documented', + detail: + 'A Knowledge table can start empty with custom columns, and Tool steps insert, update, delete, and search rows. CSV uploads are capped at 50,000 rows.', + shortValue: 'Knowledge tables agents read and write', + confidence: 'estimated', + sources: [ + { + url: 'https://relevanceai.com/docs/build/knowledge/create-knowledge', + label: 'Create and Organize Knowledge | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/admin/system-limits', + label: 'System quotas | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + richTextEditor: { + value: 'Not publicly documented', + detail: + 'No in-product rich-text document editor is described; Chat includes Slide Builder and Website Builder agents that generate content instead.', + shortValue: 'Not publicly documented', + confidence: 'unknown', + sources: [], + }, + subWorkflows: { + value: + 'Yes: a saved Tool can be attached to many agents and chained to other Tools, and Workforces hand tasks between saved agents; the same agent can join several Workforces', + detail: + 'Tools exist independently of agents in a Workforce and can be connected to multiple agents or to other tools in sequence. The older Sub-agents setting is marked legacy and may be deprecated.', + shortValue: 'Reusable Tools and agents across Workforces', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/get-started/core-concepts/workforces', + label: 'Workforces | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/build/workforces/build-an-ai-workforce/introduction-to-nodes', + label: 'Introduction to Nodes | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + customBlocks: { + value: + 'Partial: Tools are named, reusable units with defined inputs and outputs that any agent in the project can use, and can be listed in the Marketplace; cloning copies the full Tool rather than exposing it as an encapsulated block', + detail: + 'Marketplace and template sharing let others clone an agent, Tool, or Workforce into their own project, including its configuration, which is the opposite of hiding the internals behind an interface.', + shortValue: 'Reusable Tools; shared by cloning, not encapsulated', + confidence: 'estimated', + sources: [ + { + url: 'https://relevanceai.com/docs/get-started/core-concepts/tools', + label: 'Tools | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/get-started/marketplace/introduction', + label: 'Marketplace | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + }, + aiCapabilities: { + multiLlmSupport: { + value: + 'Yes: models from OpenAI, Anthropic, Google, Azure, and OpenRouter, selectable per agent, tool step, or Workforce condition, with reasoning-effort and temperature settings', + detail: + 'The changelog shows models added within weeks of release, such as Claude Sonnet 5.5 and Haiku 5.5 and Gemini 3.5 Flash in 2026, plus decision models (TypeSafe Jev 1.13, GPT-6 Luna Decisions) for Workforce conditions.', + shortValue: 'OpenAI, Anthropic, Google, Azure, OpenRouter models', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/build/agents/build-your-agent/agent-settings/language-model', + label: 'Language Model | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/changelog', + label: 'Changelog | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + agentReasoningBlocks: { + value: + 'Yes: each agent reasons over its prompt and decides which of its tools to call, and in a Workforce can decide whether to hand work to another agent over an AI connection', + detail: + 'Workforce connections are either AI connections, where the agent chooses when to pass control, or Next step connections that always fire when the previous agent finishes.', + shortValue: 'Agents choose tools and handoffs', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/get-started/core-concepts/workforces', + label: 'Workforces | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + naturalLanguageBuilding: { + value: + 'Yes: Invent builds Agents, Tools, Triggers, Knowledge tables, Workforces, and Evals from a plain-language description, and coding agents such as Claude Code can build them over MCP', + detail: + 'Invent asks before restoring versions and defaults delete and publish actions to Always ask. It is on by default in new projects, and organization admins can turn it off per project.', + shortValue: 'Invent builds agents and Workforces from plain language', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/build/invent/invent', + label: 'Build with Invent | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + knowledgeBaseRag: { + value: + 'Yes: Knowledge tables built from uploads (CSV, PDF, Excel, JSON, audio), websites, or synced Google Drive, SharePoint, Notion, and Confluence sources, searched by semantic or hybrid retrieval', + detail: + "An agent can either receive the whole knowledge set in its prompt or search it with RAG. Relevance AI's features page lists hybrid retrieval (vector, keyword/BM25, and ensemble) and configurable character or semantic chunking.", + shortValue: 'Knowledge tables with semantic/hybrid search and synced sources', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/get-started/core-concepts/knowledge', + label: 'Knowledge | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/features', + label: 'Relevance AI Features', + asOf: '2026-11-19', + }, + ], + }, + mcpSupport: { + value: + 'Yes: agents can connect preset (Notion, Canva, Atlassian) or custom remote MCP servers over Streamable HTTP; local MCP servers are not supported', + detail: + 'The same MCP server URL can be connected several times with different credentials, and each tool is labeled with its connection name.', + shortValue: 'Remote MCP servers; no local servers', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/integrations/mcp/mcp-client', + label: 'MCP client | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + evaluationGuardrails: { + value: + 'Yes on Enterprise: Evals runs simulated-conversation Test sets scored by reusable Checks, can block publishing below a minimum pass rate, and samples live conversations on Monitor dashboards', + detail: + 'Agent Evaluations are marked Enterprise-only on the pricing comparison. For guardrails, customers using their own AWS Bedrock account can apply Bedrock Guardrails to Anthropic models, and Enterprise customers can enable prompt-injection detection recorded on OpenTelemetry traces.', + shortValue: 'Pre-publish evals and live monitoring (Enterprise)', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/build/agents/evals/introduction', + label: 'Introduction to Evals | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/get-started/pricing', + label: 'Pricing | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/admin/security', + label: 'Security overview | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + humanInTheLoop: { + value: + 'Yes: each Workforce connection can run automatically, require approval, or let the agent decide; pending tasks wait for approval in the task view, and approvals can be handled in Slack', + detail: + 'Agents can also escalate to a person through escalation tools. The Tasks page groups Approvals, Escalated, and Errors views, and alarms can notify email, Slack, or Teams when too many tasks wait on approval.', + shortValue: 'Per-connection approval modes with Slack approvals', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/build/workforces/workforce-features/approvals-and-escalations', + label: 'Approvals and Escalations | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/changelog', + label: 'Changelog | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + generativeMedia: { + value: + 'Partial: image generation (a "Text to image generation" step and an Image Generator agent in Chat, 24+ image models per the features page) and voice through Phone Agents; no video generation is documented', + detail: + "Audio and video can be transcribed to text with dedicated Tool steps. OpenAI's GPT Image 1.5 was added to the image step in February 2026.", + shortValue: 'Image generation and phone voice; no video', + confidence: 'estimated', + sources: [ + { + url: 'https://relevanceai.com/features', + label: 'Relevance AI Features', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/get-started/chat/chat-agents/image-generator', + label: 'Image Generator | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/changelog', + label: 'Changelog | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + dynamicToolUse: { + value: + 'Yes: an agent picks from the tools, MCP tools, and agents connected to it at run time, and can call several in parallel when Parallel Tool Calls (beta) is on', + detail: + 'Tools are pre-attached to an agent or Workforce node; the agent chooses among them per step rather than browsing an open catalog.', + shortValue: 'Agent chooses among its attached tools', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/build/agents/build-your-agent/agent-settings/language-model', + label: 'Language Model | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + modelFallback: { + value: + 'Yes: a fallback model can be set per agent and retries a failed task once if the primary model errors or does not respond', + detail: + 'If the fallback model also fails, the task fails and follows the configured error handling.', + shortValue: 'One-time retry on a fallback model', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/build/agents/build-your-agent/agent-settings/language-model', + label: 'Language Model | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + agentSkills: { + value: + 'Partial: Snippets store shared text reused across a workspace, and Chat has saved prompts; there is no documented named skill package attached to many agents', + detail: + "Relevance AI's 'agent skills' repository is for coding assistants such as Claude Code, not a feature agents inside Relevance AI use.", + shortValue: 'Shared Snippets and saved prompts', + confidence: 'estimated', + sources: [ + { + url: 'https://relevanceai.com/docs/build/knowledge/use-snippets-for-quick-access', + label: 'Use Snippets for Quick Access | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + nativeChatDeployment: { + value: + 'Yes: Relevance Chat (web, desktop, iOS, Android), a public Chat UI link or iframe, and an embeddable chat widget', + detail: + 'Enabling a Chat UI, widget, or Task UI makes the agent publicly accessible, and usage is billed to the owner. User Level Authentication is not supported in embedded widgets.', + shortValue: 'Hosted Chat app, share link, and embeddable widget', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/build/agents/share-your-agent', + label: 'Share Your Agent | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/get-started/chat/introduction', + label: 'Chat | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + kbChunkVisibility: { + value: + 'Partial: chunking is configurable when importing documents and Knowledge is stored as tables you can open, but no retrieval debugger showing which chunks a search returned is documented', + detail: + 'The docs recommend balancing chunk size and overlap, which implies chunk-level storage, but how chunks appear in the table is not described.', + shortValue: 'Configurable chunking; no retrieval inspector documented', + confidence: 'estimated', + sources: [ + { + url: 'https://relevanceai.com/docs/build/knowledge/create-knowledge', + label: 'Create and Organize Knowledge | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + parallelExecution: { + value: + 'Yes: Parallel Tool Calls (beta) lets an agent run several tools or agent handoffs at once, and Tool loops can process list items in parallel', + detail: + 'In a Workforce, outbound connections must use the "Create new task" behavior to run in parallel.', + shortValue: 'Parallel tool calls (beta) and parallel loops', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/build/agents/build-your-agent/agent-settings/language-model', + label: 'Language Model | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/build/tools/customize-tool/looping', + label: 'Looping | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + a2aProtocol: { + value: 'Not publicly documented', + detail: + 'Agent-to-agent handoffs inside a Workforce are a Relevance AI feature, but no support for the open Agent2Agent (A2A) protocol is documented.', + shortValue: 'Not publicly documented', + confidence: 'unknown', + sources: [], + }, + loopIteration: { + value: + 'Yes: a Loop step inside a Tool iterates over a list, one item at a time or all in parallel, with {{foreach.item}} and {{foreach.index}} variables', + detail: + 'By default an item that errors is skipped and shown in the Loop’s Errors tab; the Loop can be set to throw instead.', + shortValue: 'Loop step over lists, sequential or parallel', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/build/tools/customize-tool/looping', + label: 'Looping | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + }, + integrations: { + integrationCount: { + value: + '2,000+ integrations per Relevance AI (its catalog listed 2,728 on 2026-11-19), most added through a Pipedream connection; the docs separately cite 1,000+ app triggers', + detail: + "The integrations docs split the catalog into native integrations built by Relevance AI and Pipedream integrations that 'extend the catalog to 2,000+ services'. The enterprise page and llms.txt instead say 1,000+ native connectors, so Relevance AI's own pages give two different figures.", + shortValue: '2,000+ (most via Pipedream)', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/integrations', + label: 'Relevance AI Integrations', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/integrations/introduction', + label: 'Integrations | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/enterprise', + label: 'Relevance AI Enterprise', + asOf: '2026-11-19', + }, + ], + }, + triggerTypes: { + value: + 'Integration triggers (Gmail, Slack, Teams, Google Drive, and more), schedules and cron, webhooks, API, SDK, Tools as triggers, manual chat, and bulk runs over a CSV', + detail: + 'Premium triggers (WhatsApp, LinkedIn, Telegram) are on all paid plans; Enterprise triggers (Salesforce, Snowflake, Zendesk) require Enterprise. Triggers can be paused and resumed, and a Trigger status page alerts on failures.', + shortValue: 'App events, schedules, webhooks, API, SDK, bulk', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/build/agents/build-your-agent/triggers', + label: 'Triggers | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/get-started/pricing', + label: 'Pricing | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + customCodeSteps: { + value: + 'Yes: Python and JavaScript code steps inside Tools, plus an API step for custom HTTP calls', + detail: + 'The features page lists sandboxed code execution in Python and JavaScript; the Python step docs cover PyPI packages and helper functions.', + shortValue: 'Python and JavaScript steps, API step', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/build/tools/tool-steps/python-code/code-python', + label: 'Python Tool step | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/features', + label: 'Relevance AI Features', + asOf: '2026-11-19', + }, + ], + }, + codeSandboxRuntime: { + value: + 'Partial: a Python step can declare any PyPI package in its "PyPI Packages" list, but OS-level packages and extra CLI binaries are not documented', + detail: 'No package option for the JavaScript step is documented.', + shortValue: 'PyPI packages per Python step; no OS packages', + confidence: 'estimated', + sources: [ + { + url: 'https://relevanceai.com/docs/build/tools/tool-steps/python-code/code-python', + label: 'Python Tool step | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + apiPublishing: { + value: + 'Yes: an API trigger starts an agent with an HTTP POST, and Tools can be run over the API; results are sent to a configured destination rather than returned in the response', + detail: + "The docs say the API is 'officially supported only for triggering Agents and Tools'. The JavaScript SDK wraps the same endpoint with task events and streaming.", + shortValue: 'HTTP trigger for agents and tools', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/build/agents/build-your-agent/agent-triggers/api-trigger', + label: 'API Triggers | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + extensibilitySdk: { + value: + 'An official JavaScript/TypeScript SDK (@relevanceai/sdk), an MCP server, a Claude Code plugin and agent-skills repository, custom API keys for unlisted services, and a Marketplace where vetted builders sell agents and tools', + detail: + 'The SDK runs in Node.js, Deno, Bun, Cloudflare Workers, and the browser. No Python SDK or developer kit for building new native integrations is documented.', + shortValue: 'JS/TS SDK, MCP server, Claude Code plugin, Marketplace', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/sdk/introduction', + label: 'JavaScript SDK | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/get-started/core-concepts/mcp-plugins', + label: 'MCP & Plugins | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + mcpPublishing: { + value: + 'Yes, at project level: the Relevance AI MCP server (mcp.relevanceai.com) exposes a project’s agents, tools, and knowledge to Claude, ChatGPT, Cursor, VS Code, and other MCP clients', + detail: + 'Users with Viewer or Chat roles connect in a restricted mode that runs existing agents without building or editing them. The server exposes the project as a whole rather than publishing a single workflow as its own MCP server.', + shortValue: 'Project-wide MCP server for external clients', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/integrations/mcp/mcp-server', + label: 'MCP server | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + }, + pricing: { + pricingModel: { + value: + 'Organization-level subscription with two meters: Actions (one per Tool run, even if it fails) and Vendor Credits (model and tool costs passed through without markup); Enterprise is sales-led', + detail: + 'Plan Actions reset at each renewal while Vendor Credits roll over while subscribed. Top-ups cost $80 per 1,000 Actions and $20 per 10,000 Vendor Credits. On 2026-11-19 the relevanceai.com/pricing page showed only the Enterprise tier; self-serve prices are on the docs pricing page.', + shortValue: 'Subscription plus Actions and Vendor Credits', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/admin/subscriptions/plans', + label: 'Plans and credits | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/pricing', + label: 'Relevance AI Pricing', + asOf: '2026-11-19', + }, + ], + }, + entryPaidPlan: { + value: + 'Pro: $19/month billed annually or $29/month billed monthly, with 2,500 Actions and 10,000 Vendor Credits ($20) a month, 2 build users, and 1 project', + detail: + 'Team is $234/month billed annually or $349/month billed monthly, with 7,000 Actions and 35,000 Vendor Credits a month, 5 build users, 45 end users, and 5 shared projects. Prices are from the docs pricing page; the marketing pricing page listed only Enterprise when checked.', + shortValue: '$19/mo annual or $29/mo monthly (Pro)', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/get-started/pricing', + label: 'Pricing | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + freeTier: { + value: 'No: the Free plan is retired for new signups and existing accounts', + detail: 'Organizations that were on Free must upgrade to Pro or Team to keep using them.', + shortValue: 'No free plan', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/get-started/pricing', + label: 'Pricing | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + byok: { + value: + 'Yes, on every plan: your own model keys bypass Vendor Credits (with credits kept as a fallback), but each Tool run still counts as an Action', + detail: + 'Supported connections include an OpenAI key, AWS Bedrock or Google Cloud for Anthropic models, Google AI Studio or Google Cloud for Gemini, Azure, and OpenRouter. Bedrock role assumption requires Enterprise.', + shortValue: 'BYOK on all plans; Actions still apply', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/admin/subscriptions/plans', + label: 'Plans and credits | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/integrations/llm-integrations/anthropic', + label: 'Anthropic LLM models | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + }, + security: { + soc2: { + value: 'Yes: SOC 2 Type II compliant', + detail: + 'Compliance documentation and third-party assessment reports are shared with Enterprise customers on request under NDA.', + shortValue: 'SOC 2 Type II', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/admin/security', + label: 'Security overview | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + dataResidency: { + value: 'Yes: US (N. Virginia), EU (London), or AU (Sydney), chosen at signup', + detail: + 'The region cannot be changed after the organization is created without contacting support. Relevance AI also says it hosts some OpenAI and open-source models in its own multi-region AWS and Azure environments.', + shortValue: 'US, EU, or AU region at signup', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/admin/security', + label: 'Security overview | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + rbac: { + value: + 'Enterprise only, still rolling out: organization roles (Owner, Admin, Member, Viewer), project roles (Admin, Editor, Member, Chat, Viewer), and asset roles (Admin, Member, Viewer)', + detail: + 'IdP groups from Entra ID or Okta can be synced through WorkOS about every 60 minutes and assigned to projects and assets.', + shortValue: 'Org, project, and asset roles (Enterprise)', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/enterprise/rbac', + label: 'Role-based access controls (RBAC) | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/enterprise/rbac-groups', + label: 'Directory Sync and Groups | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + auditLogging: { + value: + 'Enterprise only: audit logs and execution traces stream in OpenTelemetry format to your own S3 bucket; organization Owners, Admins, and Viewers can view audit logs', + detail: + 'S3 is the only supported destination, and the OpenTelemetry stream requires RBAC to be enabled. The features page lists user, IP, device, and timestamp on audit events.', + shortValue: 'Enterprise audit logs streamed to S3', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/enterprise/streaming-events', + label: 'Event Streaming for Observability (OpenTelemetry) | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/get-started/pricing', + label: 'Pricing | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/enterprise/rbac', + label: 'Role-based access controls (RBAC) | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/changelog', + label: 'Changelog | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + additionalCompliance: { + value: + 'SOC 2 Type II and GDPR; no ISO 27001, HIPAA, PCI, or FedRAMP attestation is documented', + detail: + 'The data-retention docs say retention policies help support frameworks such as HIPAA, but no HIPAA attestation or BAA is described. The Vanta-hosted Trust Center at trust.relevanceai.com renders client-side and could not be read directly.', + shortValue: 'SOC 2 Type II, GDPR', + confidence: 'estimated', + sources: [ + { + url: 'https://relevanceai.com/docs/admin/security', + label: 'Security overview | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + modelAndToolGovernance: { + value: + 'Yes on Enterprise: model controls block providers or single models and keep new models off until approved, and integration controls hide integrations that are not enabled', + detail: + 'Before a model is turned off, admins see which platform features, Agents, and LLM tool steps it affects.', + shortValue: 'Model and integration allowlists (Enterprise)', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/enterprise/model-controls', + label: 'Model controls | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/enterprise/integration-controls', + label: 'Organization-level integration controls | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + credentialGovernance: { + value: + 'Partial, Enterprise and rolling out: users choose whether their OAuth accounts are shared on an asset, and User Level Authentication makes shared agents use each person’s own accounts', + detail: + 'Custom API keys stored in a project can be used by any builder in that project, which the docs flag as a shared-access risk.', + shortValue: 'Per-asset OAuth sharing and per-user auth (Enterprise)', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/enterprise/asset-controls', + label: 'Asset-level authentication controls | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/enterprise/user-level-authentication', + label: 'User Level Authentication | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + whiteLabeling: { + value: + 'Partial: the embeddable chat widget can hide Relevance AI branding and set a primary color; full white-labeling of the app is not documented', + detail: 'Widget options include hide_logo, primary_color, and bubble icon settings.', + shortValue: 'Chat widget branding can be hidden', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/build/agents/share-your-agent', + label: 'Share Your Agent | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + dataRetention: { + value: + 'Yes on Enterprise: admins set retention windows for tool runs, agent conversations, Workforce tasks, and old versions; Pro and Team list 90-day task history', + detail: + 'Data Retention must be enabled by the Relevance AI team. The security page says run logs on paid tiers are otherwise stored until you delete them.', + shortValue: 'Configurable retention (Enterprise); 90-day history on Pro/Team', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/enterprise/data-retention', + label: 'Data Retention | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/get-started/pricing', + label: 'Pricing | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/admin/security', + label: 'Security overview | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + piiRedaction: { + value: + 'Partial, Enterprise: Visual Data Masking (limited beta) hides emails, phone numbers, card numbers, and names in the UI only, and PII redaction can be enabled on streamed logs', + detail: + 'Masking is display-only: agents still process the unmasked data, and the docs warn the detection model can miss sensitive information. The features page also lists PII masking before data reaches the model; Bedrock Guardrails can mask PII for BYO Anthropic models.', + shortValue: 'UI-only masking (beta) and log redaction', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/build/agents/build-your-agent/agent-settings/visual-data-masking', + label: 'Visual Data Masking | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/enterprise/streaming-events', + label: 'Event Streaming for Observability (OpenTelemetry) | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + sso: { + value: + 'Yes on Enterprise: SAML 2.0 and OIDC SSO through WorkOS (Okta, Entra ID, Google Workspace, Auth0, and more), plus directory sync of IdP groups', + detail: + 'SSO is set up with a Relevance AI representative. Email is the account identifier, so an email change in the IdP creates a new account rather than updating the existing one.', + shortValue: 'SAML/OIDC SSO with directory sync (Enterprise)', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/enterprise/sso-setup', + label: 'Single Sign-On (SSO) Setup | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + sessionPolicy: { + value: 'Not publicly documented', + detail: + 'No admin-configurable session lifetime or idle timeout is described in the SSO, RBAC, or security docs.', + shortValue: 'Not publicly documented', + confidence: 'unknown', + sources: [], + }, + thirdPartyVetting: { + value: + 'Mixed: native integrations are built by Relevance AI, most of the catalog comes through Pipedream, Marketplace listings are reviewed by the Relevance AI team, and agents can connect any remote MCP server URL', + detail: + 'Only verified Relevance Builders can submit to the Marketplace, and each submission and update is reviewed before it goes live. No review process is described for custom MCP server URLs.', + shortValue: 'Reviewed Marketplace; open custom MCP URLs', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/get-started/marketplace/relevance-builders/submit-agents', + label: 'Submit Agents to the Marketplace | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/integrations/introduction', + label: 'Integrations | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + }, + observability: { + tracingDepth: { + value: + 'Each task shows its steps, tool runs, and credit cost per tool and LLM call; Analytics dashboards (Team and up) show usage, errors, and concurrency; Enterprise can stream execution traces via OpenTelemetry', + detail: + 'The Tasks page lists Approvals, Escalated, and Errors views with day-by-day timelines, and the Home page shows a 14-day task health bar.', + shortValue: 'Per-task step and cost view; dashboards; OTEL on Enterprise', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/build/agents/give-your-agent-tasks/tasks-page', + label: 'Task Operations | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/enterprise/analytics', + label: 'Analytics | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/changelog', + label: 'Changelog | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + durabilityModel: { + value: + 'Errored tasks can be rerun from the failing step, a fallback model retries a failed task once, and the features page lists durable execution, checkpointing, exponential-backoff retries, and a dead-letter queue', + detail: + 'Task statuses include Exhausted retries and Queued for rerun. The detailed retry policy is described on the features page rather than in the docs.', + shortValue: 'Rerun from failing step; retries and checkpointing', + confidence: 'estimated', + sources: [ + { + url: 'https://relevanceai.com/docs/build/agents/give-your-agent-tasks/tasks-page', + label: 'Task Operations | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/features', + label: 'Relevance AI Features', + asOf: '2026-11-19', + }, + ], + }, + failureAlerting: { + value: + 'Yes, on paid plans: alert rules on task status (timed out, unrecoverable error, exhausted retries, and more), tool errors, and publishes, sent to email, Slack, or Teams', + detail: + 'Rate-based alarms fire when the share of failed, approval-waiting, or escalated tasks crosses a threshold, and a Trigger status page alarms on trigger failures. Usage alerts email admins on credit and Action consumption.', + shortValue: 'Email, Slack, or Teams alerts on failures', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/build/agents/build-your-agent/alerts', + label: 'Alerts | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/changelog', + label: 'Changelog | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + dataDrains: { + value: + 'Yes on Enterprise: audit logs and execution traces stream in OpenTelemetry format to a customer S3 bucket; the features page also lists Delta Sharing export', + detail: + 'Direct OpenTelemetry collector endpoints are on the roadmap. Data can also be exported manually as CSV, Excel, or JSON.', + shortValue: 'OTEL export to S3 (Enterprise)', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/enterprise/streaming-events', + label: 'Event Streaming for Observability (OpenTelemetry) | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/admin/security', + label: 'Security overview | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + asyncExecution: { + value: + 'Yes: API and webhook triggers start a task and return without the result, which is delivered to a configured destination; the SDK can subscribe to task status events', + detail: + 'When the concurrency limit is reached, additional tasks are queued until capacity frees up.', + shortValue: 'Fire-and-forget triggers; SDK task events', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/build/agents/build-your-agent/agent-triggers/api-trigger', + label: 'API Triggers | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/sdk/introduction', + label: 'JavaScript SDK | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + executionLimits: { + value: + 'Tools run up to 15 minutes (or 24 hours for long-running Tools), agents up to 15 minutes after each user action, and bulk runs up to 24 hours; concurrency is set by plan, with no published numbers', + detail: + 'Uploads are capped at 100 MB and CSV uploads at 50,000 rows. The pricing table lists concurrent agent tasks as Standard on Pro, More on Team, and Custom on Enterprise.', + shortValue: '15-minute agent steps, 24-hour bulk runs', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/admin/system-limits', + label: 'System quotas | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + partialFailureHandling: { + value: + 'Partial: a Loop skips items that error by default, and a fallback model can retry a failed task, but an error in a branch or standard Tool step stops the Tool', + detail: 'Legacy Sub-agent connections allowed custom error-handling instructions.', + shortValue: 'Loops skip failed items; other step errors stop the Tool', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/build/tools/customize-tool/looping', + label: 'Looping | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/build/tools/customize-tool/branching', + label: 'Branching | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + unattendedExecution: { + value: + "Yes: scheduled, integration, webhook, and API triggers run agents on Relevance AI's hosted infrastructure with no client device involved", + detail: 'The mobile and desktop apps are clients for Chat and monitoring, not runtimes.', + shortValue: 'Runs server-side on hosted infrastructure', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/build/agents/build-your-agent/triggers', + label: 'Triggers | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + }, + support: { + supportChannels: { + value: + 'Community forum, in-app chat tickets, and email on all plans; priority support on Team; a dedicated Slack channel, account manager, and staffed calls on Enterprise', + detail: + 'Free-to-Team customers can request calls with Harley, an AI support agent in beta, during Sydney business hours.', + shortValue: 'Forum, chat, email; Slack and AM on Enterprise', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/get-started/support', + label: 'How to reach out to our support team | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + sla: { + value: + 'Relevance AI lists a 99.9% uptime SLA on its enterprise page; the plans it covers are not stated', + detail: 'No SLA terms or credits are published.', + shortValue: '99.9% uptime SLA (plan not stated)', + confidence: 'estimated', + sources: [ + { + url: 'https://relevanceai.com/enterprise', + label: 'Relevance AI Enterprise', + asOf: '2026-11-19', + }, + ], + }, + community: { + value: + 'A peer-to-peer community forum (community.relevanceai.com) with Relevance AI staff and partners, plus a partner directory', + detail: 'Relevance AI notes the forum is primarily monitored by community members.', + shortValue: 'Community forum and partner network', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/docs/get-started/support', + label: 'How to reach out to our support team | Relevance AI Docs', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/', + label: 'Relevance AI homepage', + asOf: '2026-11-19', + }, + ], + }, + companyMaturity: { + value: + 'OnSearch Pty Ltd, trading as Relevance AI, with offices in Sydney and San Francisco. Raised a $24M Series B in May 2025 led by Bessemer Venture Partners, with King River Capital, Insight Partners, and Peak XV; the team was over 80 people at the time', + detail: + 'Customers named on its site include Canva, Autodesk, Databricks, Qualified, and SafetyCulture. As a venture-backed private company, it carries more switching risk than a large public vendor.', + shortValue: '$24M Series B (2025), Sydney and San Francisco', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/blog/the-ai-workforce-revolution-24m-series-b-to-accelerate-our-mission', + label: 'The AI Workforce Revolution: $24M Series B | Relevance AI Blog', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/about-us', + label: 'About us | Relevance AI', + asOf: '2026-11-19', + }, + ], + }, + academy: { + value: + 'Yes: Relevance AI Academy offers free self-paced courses, and an AI Ops Bootcamp offers certification', + detail: 'The docs point new builders to the Academy for structured, step-by-step courses.', + shortValue: 'Free Academy courses and AI Ops certification', + confidence: 'verified', + sources: [ + { + url: 'https://relevanceai.com/', + label: 'Relevance AI homepage (Resources menu)', + asOf: '2026-11-19', + }, + { + url: 'https://relevanceai.com/docs/build/agents/build-your-agent/triggers', + label: 'Triggers | Relevance AI Docs', + asOf: '2026-11-19', + }, + ], + }, + }, + }, +} diff --git a/apps/sim/lib/compare/data/index.ts b/apps/sim/lib/compare/data/index.ts index fee93aab365..76691490e4f 100644 --- a/apps/sim/lib/compare/data/index.ts +++ b/apps/sim/lib/compare/data/index.ts @@ -1,5 +1,6 @@ export { claudeCoworkProfile } from '@/lib/compare/data/competitors/claude-cowork' export { crewaiProfile } from '@/lib/compare/data/competitors/crewai' +export { difyProfile } from '@/lib/compare/data/competitors/dify' export { dustProfile } from '@/lib/compare/data/competitors/dust' export { flowiseProfile } from '@/lib/compare/data/competitors/flowise' export { gumloopProfile } from '@/lib/compare/data/competitors/gumloop' @@ -12,6 +13,7 @@ export { openaiAgentkitProfile } from '@/lib/compare/data/competitors/openai-age export { openClawProfile } from '@/lib/compare/data/competitors/openclaw' export { pipedreamProfile } from '@/lib/compare/data/competitors/pipedream' export { powerAutomateProfile } from '@/lib/compare/data/competitors/power-automate' +export { relevanceAiProfile } from '@/lib/compare/data/competitors/relevance-ai' export { retoolProfile } from '@/lib/compare/data/competitors/retool' export { stackaiProfile } from '@/lib/compare/data/competitors/stackai' export { tinesProfile } from '@/lib/compare/data/competitors/tines' diff --git a/apps/sim/scripts/test-landing-seo-e2e.ts b/apps/sim/scripts/test-landing-seo-e2e.ts index bc4a0af9ee6..f241ff5f18d 100644 --- a/apps/sim/scripts/test-landing-seo-e2e.ts +++ b/apps/sim/scripts/test-landing-seo-e2e.ts @@ -13,6 +13,7 @@ import { SIM_SITE_URL } from '@sim/utils/site' import { load } from 'cheerio' import matter from 'gray-matter' import { + difyProfile, dustProfile, gumloopProfile, langflowProfile, @@ -22,6 +23,7 @@ import { openClawProfile, type Prose, pipedreamProfile, + relevanceAiProfile, vellumProfile, } from '@/lib/compare/data' import { ContentFrontmatterSchema } from '@/lib/content/schema' @@ -494,6 +496,8 @@ async function run() { pipedreamProfile, openClawProfile, openaiAgentkitProfile, + difyProfile, + relevanceAiProfile, ]) { const { status, body } = await get(`/comparisons/${profile.id}`) if (status !== 200) { From ef4b9370935870171d10812bb983a5418e1b4090 Mon Sep 17 00:00:00 2001 From: Waleed Latif Date: Thu, 19 Nov 2026 20:04:20 -0800 Subject: [PATCH 2/3] chore(docs): sync docs icons with the new comparison brand icons --- apps/docs/components/icons.tsx | 34 ++++++++++++++++++++++++++++++++++ 1 file changed, 34 insertions(+) diff --git a/apps/docs/components/icons.tsx b/apps/docs/components/icons.tsx index 190e8091440..05b17279bbd 100644 --- a/apps/docs/components/icons.tsx +++ b/apps/docs/components/icons.tsx @@ -1828,6 +1828,40 @@ export function DustIcon(props: SVGProps) { ) } +export function DifyIcon(props: SVGProps) { + return ( + + + + + ) +} + +export function RelevanceAIIcon(props: SVGProps) { + return ( + + + + + + ) +} + export function OpenClawIcon(props: SVGProps) { const id = useId() const gid = (name: string) => `openclaw_${name}_${id}` From a7dc9179a49e7dc5cca30aa630488e91e5015498 Mon Sep 17 00:00:00 2001 From: Waleed Latif Date: Thu, 19 Nov 2026 20:06:04 -0800 Subject: [PATCH 3/3] fix(comparisons): read comparison leads as text in the seo e2e and tighten relevance ai sourcing --- .../sim/lib/compare/data/competitors/relevance-ai.ts | 6 +++--- apps/sim/scripts/test-landing-seo-e2e.ts | 12 +++++++++--- 2 files changed, 12 insertions(+), 6 deletions(-) diff --git a/apps/sim/lib/compare/data/competitors/relevance-ai.ts b/apps/sim/lib/compare/data/competitors/relevance-ai.ts index 83433b0defc..83ee20440e3 100644 --- a/apps/sim/lib/compare/data/competitors/relevance-ai.ts +++ b/apps/sim/lib/compare/data/competitors/relevance-ai.ts @@ -48,7 +48,7 @@ export const relevanceAiProfile: CompetitorProfile = { { question: 'Can I self-host Relevance AI, or do I need a platform like Sim?', answer: - "Relevance AI cannot be self-hosted: it runs as a multi-tenant hosted service, its security page says single-tenant options are still in the works, and its code is not published. Sim's core is Apache 2.0 licensed and can be self-hosted with Docker Compose or Kubernetes (Helm), so your team runs the whole platform. Relevance AI's answer to data location is a region choice at signup (US, EU, or Australia), which cannot be changed later.", + "Relevance AI cannot be self-hosted: it runs as a multi-tenant hosted service, its security page says single-tenant options are still in the works, and its code is not published. Sim's core is Apache 2.0 licensed and can be self-hosted with Docker Compose or Kubernetes (Helm), so your team runs the whole platform. Relevance AI's answer to data location is a region choice at signup (US, EU, or Australia); changing it afterwards means contacting support.", }, { question: 'How does Relevance AI pricing compare to Sim?', @@ -154,8 +154,8 @@ export const relevanceAiProfile: CompetitorProfile = { shortDescription: 'SSO, RBAC, audit logs, and evals need Enterprise; RBAC is still rolling out.', source: { - url: 'https://relevanceai.com/docs/get-started/pricing', - label: 'Pricing | Relevance AI Docs', + url: 'https://relevanceai.com/docs/enterprise/rbac', + label: 'Role-based access controls (RBAC) | Relevance AI Docs', asOf: '2026-11-19', }, }, diff --git a/apps/sim/scripts/test-landing-seo-e2e.ts b/apps/sim/scripts/test-landing-seo-e2e.ts index f241ff5f18d..39ccab3f8fb 100644 --- a/apps/sim/scripts/test-landing-seo-e2e.ts +++ b/apps/sim/scripts/test-landing-seo-e2e.ts @@ -505,9 +505,15 @@ async function run() { continue } const text = decodeHtml(body) - const lead = profile.leadAnswer?.map((part) => (typeof part === 'string' ? part : part.text)) - const firstSection = text.indexOf(' (typeof part === 'string' ? part : part.text)) + .join('') + const $ = load(body) + const blocks = $('p, h2').toArray() + const leadAt = lead + ? blocks.findIndex((el) => el.tagName === 'p' && $(el).text().includes(lead)) + : -1 + const firstSection = blocks.findIndex((el) => el.tagName === 'h2') if (leadAt < 0 || leadAt > firstSection) { failures.push(`${profile.id}: lead answer missing or below the first section`) }