Dify
Dify is an open-source platform for building, deploying, and operating LLM applications, agents, RAG pipelines, and workflows through a visual canvas, with cloud-hosted and self-hosted deployment options.
What is Dify?
Dify is an open-source platform for building AI applications. Its official documentation describes it as a workspace for creating agents, agentic workflows, and chatbots that use your own data, then publishing them as web apps or integrating them through APIs. The platform combines a visual canvas, a built-in RAG knowledge pipeline, model routing across providers, and a plugin marketplace for external tools.
Dify is not a single chatbot product. It is a development and operations layer for teams that want to assemble LLM-powered applications with less infrastructure work than a code-first framework requires. It offers five application types: Chatbot, Agent, Text Generator, Workflow, and Chatflow, each suited to a different level of conversational or automation complexity.
How does Dify work?
A Dify application begins with a model provider. Dify is model-agnostic; in Dify Cloud, new workspaces receive message credits for trying models from OpenAI, Anthropic, Gemini, and xAI, and users can switch to their own API keys afterward. Builders then assemble a workflow or chatflow on a visual canvas using nodes—individual steps such as user input, retrieval, branching, tool calls, or output.
For knowledge-grounded tasks, the Knowledge feature handles retrieval-augmented generation: users can upload local files, sync Notion workspaces, or crawl webpages, and Dify processes the data for context-aware answers. Workflows can also be triggered automatically by schedules, external events, or webhooks, subject to plan-level quotas. The finished application can be published as a web app or called through Dify's API.
Use Cases
- Building an internal knowledge-base assistant over uploaded documents or synced Notion pages
- Automating a multi-step business process, such as routing incoming requests and retrieving reference data before generating a response
- Creating a customer-facing chatbot with controlled conversation flow and tool access
- Prototyping and deploying an agent that can choose actions and call external tools where the plan and configuration allow
- Publishing a RAG pipeline as a reusable knowledge workflow for other applications
Pricing
Pricing model: Freemium
| Plan | Price | Billing | Included usage | Notes |
|---|---|---|---|---|
| Sandbox | Free | — | 200 message credits (one-time allocation, not renewed monthly), 1 workspace, 1 member, 5 apps, 50 knowledge documents, 50MB knowledge storage, 10 knowledge requests/min, up to 2 triggers per workflow, 3,000 trigger events. | — |
| Professional | $59 | per month | 5,000 message credits/month, 3 team members, 50 apps, 500 knowledge documents, 5GB knowledge storage, 100 knowledge requests/min, unlimited triggers per workflow, 20,000 trigger events/month, no Dify API rate limit. $590/year when billed annually. | — |
| Team | $159 | per month | 10,000 message credits/month, 50 team members, 200 apps, 1,000 knowledge documents, 20GB knowledge storage, 1,000 knowledge requests/min, unlimited triggers and trigger events, priority workflow execution. $1,590/year when billed annually. | — |
| Community (Self-hosted) | Free | — | Core platform features under a modified Apache-2.0 license, single workspace, self-managed infrastructure. Commercial use is permitted with conditions; redistributing as a competing SaaS platform is restricted. | — |
Common Limitations
- Knowledge Pipeline publishing · Publishing a knowledge base as a reusable pipeline is not available on Sandbox; it requires Professional or Team.
- Trigger count per workflow · Sandbox accounts are limited to 2 triggers per workflow. Higher trigger counts are available on Professional and Team.
- Message credits · Sandbox includes 200 message credits as a one-time allocation that does not renew monthly. Professional and Team credits are monthly and consumed based on model type; once exhausted, users can switch to their own API keys.
- Knowledge request rate limit · Every workspace has a per-minute knowledge base request limit determined by plan (Sandbox 10/min, Professional 100/min, Team 1,000/min). Refused requests still count toward the limit, so rapid retries keep the workspace blocked.
- Open-source license conditions · The Community Edition uses a modified Apache-2.0 license that permits commercial use but restricts redistributing Dify as a commercial SaaS platform whose value is primarily Dify itself.
- Enterprise controls · SSO, RBAC, audit logs, multi-workspace management, and negotiated SLAs are available only in the Enterprise edition.
Key Features
- Visual Workflow and Chatflow builder
- Agent construction and runtime
- RAG knowledge pipeline with file, Notion, and web ingestion
- Model-agnostic provider configuration
- Plugin and tool ecosystem
Dify vs. n8n and LangChain
This comparison focuses on documented positioning, deployment model, and plan structure. It does not rank answer quality or claim universal superiority; Dify, n8n, and LangChain are built around different primary workflows.
| Tool | Positioning | Key differences | Best fit | Pricing note |
|---|---|---|---|---|
| Dify Official product overview Official pricing (plan details) Docs: open-source platform description | An open-source LLM application development platform for building, deploying, and operating AI apps, agents, RAG pipelines, and workflows. | Combines a visual workflow/agent canvas with a built-in RAG knowledge pipeline, model-agnostic provider access, and a plugin ecosystem. Offers both cloud-hosted plans and a self-hosted Community Edition under a modified Apache-2.0 license. | Teams that want to assemble and publish LLM applications—especially knowledge-grounded assistants and multi-step AI workflows—with visual tooling and a choice between managed cloud and self-hosting. | Free Sandbox with one-time 200 message credits; Professional at $59/month ($590/year); Team at $159/month ($1,590/year). Community Edition is free to self-host under a modified Apache-2.0 license. Enterprise pricing is custom. |
| n8n Official homepage | A workflow automation platform for building agents and multi-step workflows with full visibility, built around integrations and data flow across systems. | n8n's primary design center is general workflow automation with 500+ native integrations and HTTP/MCP connectivity, where AI agents and model calls are embedded as steps within broader cross-system processes. Dify is centered on LLM application development and RAG, with knowledge management and app publishing as first-class concerns. | Automating processes that span multiple external systems—for example, triggering an AI step from a CRM event, enriching data, and writing results to another tool—rather than building a standalone RAG application. | Open-source core with self-hosting; paid n8n Cloud plans are also offered. AI model usage is separate from the automation platform layer. |
| LangChain Official homepage | A code-first agent engineering framework and platform for building, testing, deploying, monitoring, and governing agents across the development lifecycle. | LangChain provides maximum control through code and framework abstractions, plus LangSmith for tracing, evaluation, and production monitoring. Dify provides a visual application builder and managed runtime with less code, but correspondingly less low-level control over agent logic and retrieval internals. | Engineering teams that need fine-grained control over agent behavior, custom retrieval logic, or deep instrumentation, and are prepared to own more of the implementation and operational surface. | LangChain framework is open source; the LangSmith agent engineering platform has separate paid tiers. Model provider costs are separate. |
Comparison checked against official product information on 2026-11-05. Features and pricing may change.
Frequently Asked Questions
Is Dify open source?
Yes. The Community Edition is source-available under a modified Apache-2.0 license. It permits commercial use but adds conditions, including a restriction on redistributing Dify as a commercial SaaS platform that competes with Dify Cloud itself. The license is not a standard OSI Apache-2.0 grant without conditions.
What can I do on the free Sandbox plan?
Sandbox includes 200 message credits as a one-time allocation, 5 apps, 50 knowledge documents, 50MB knowledge storage, 10 knowledge requests per minute, and up to 2 triggers per workflow. Knowledge Pipeline publishing is not available on Sandbox. It is intended for evaluation, not sustained production use.
Can I self-host Dify for free?
Yes. The Community Edition can be self-hosted without a platform license fee and includes core features. Enterprise features such as SSO, RBAC, audit logs, and multi-workspace management require the Enterprise edition.
How is Dify different from n8n or LangChain?
Dify is built around LLM application development: RAG, agent runtime, prompt orchestration, and publishing AI apps. n8n is a general-purpose workflow automation platform with a very large integration library, where AI nodes are one category among many. LangChain is a code-first framework for agent development, with more control but more implementation work. They can be complementary; Dify is not a drop-in replacement for either.
Are message credits the same as model API costs?
No. Dify Cloud message credits are platform-provided allowances for trying built-in model access. Once credits are exhausted on a paid plan, users can switch to their own model provider API keys and pay those providers directly. Self-hosted Community Edition does not include platform message credits; model costs are separate.
Alternative Tools
-
LangChain LangChain is an MIT-licensed framework for building agents and LLM-powered applications, with create_agent as its current entry point and LangGraph as the underlying orchestration runtime. LangSmith, a separate paid product, adds tracing, evaluation, and deployment. -
CrewAI CrewAI is an open-source Python framework for orchestrating role-playing AI agent teams (Crews) within structured, stateful workflows (Flows), with a hosted AMP platform for enterprise governance. -
Google ADK Google ADK is an Apache 2.0 open-source framework for building, evaluating, and deploying multi-agent systems, with first-class support for Gemini and Google Cloud and coverage across Python, TypeScript, Go, Java, and Kotlin. -
OpenAI Agents SDK OpenAI Agents SDK is a lightweight, MIT-licensed Python and TypeScript framework for building multi-agent workflows using a minimal set of primitives: agents, tools, handoffs, guardrails, sessions, and tracing. -
Vercel AI SDK Vercel AI SDK is a free, Apache 2.0 TypeScript toolkit for building streaming AI UIs and agents with React, Next.js, Vue, Svelte, and Node.js. It unifies 100+ models across providers behind a single API, ships React hooks for chat and generative UI, and requires Node.js 22+ with ESM-only imports in v7. It is not a RAG framework or a full agent orchestrator. -
Browser Use Browser Use is an MIT-licensed open-source AI browser agent (Python and TypeScript) that lets LLMs control a browser, paired with a usage-based cloud offering at $0.02 per browser-hour for managed stealth browsers, CAPTCHA solving, and a hosted agent API.