Dify
Dify’s own pitch is to “build agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace”. You design apps on a visual canvas, test them in place, and publish them with an API.
What you get
- A visual workflow canvas for building and testing AI workflows.
- Hundreds of models from many providers, including any OpenAI-compatible API.
- A Prompt IDE, a RAG pipeline and LLMOps for writing prompts, retrieving from your documents and monitoring production.
- Agents with their own sandbox, using tools from the Dify Marketplace, MCP servers or your own APIs.
- Observability through Opik, Langfuse and Arize Phoenix.
- Backend-as-a-Service: every app you build comes with an API.
Self-hosting
The free Community Edition runs with Docker Compose (v2.24.0 or later) on a machine with at least 2 CPU cores and 4 GiB of RAM, and gives you one workspace. Dify Enterprise adds SSO, multiple workspaces and official support.
Dify Cloud plans (checked 1 Nov 2026 on dify.ai/pricing)
| Plan | Price | Highlights |
|---|---|---|
| Sandbox | Free | 200 message credits, 1 member, 5 apps, 50 knowledge documents (50 MB), 30 days of logs |
| Professional | $59/mo per workspace, or $590/yr | 5,000 message credits a month, 3 members, 50 apps, 500 documents (5 GB) |
| Team | $159/mo per workspace, or $1,590/yr | 10,000 message credits a month, 50 members, 200 apps, 1,000 documents (20 GB) |
| Enterprise | Custom | SSO, multiple workspaces, negotiated SLAs and support |
Prices exclude taxes.
Mind the license
Dify’s source is on GitHub under the Dify Open Source License: Apache 2.0 plus two conditions. You need Dify’s written permission to run it as a multi-tenant service (one tenant is one workspace), and you can’t remove or change the Dify logo or copyright in its console or apps when you use its frontend. A single-workspace deployment for your own team isn’t a multi-tenant service, but the branding rule still applies.
Who it’s for
- Product teams that want working LLM apps and RAG chatbots without a big engineering project.
- Mixed teams, where non-developers edit prompts and knowledge bases while developers call the API.
- Companies that need to self-host to keep data in-house.
Alternatives
Prefer to build in code? LangChain is the MIT-licensed framework most teams start with; see Dify vs LangChain. For role-based agent teams, look at CrewAI.