Dify

Dify is an open-source LLM application platform that lets developers and product teams build AI agents, agentic workflows and retrieval-augmented generation (RAG) apps in a visual editor. It sits between a code framework and a closed no-code tool: you design flows on a canvas, but you can run the whole stack on your own servers. The combination of a visual workflow builder, a full knowledge pipeline and a plugin marketplace in one self-hostable package is what draws most teams to it.

What is Dify used for?

Dify is used to turn a prompt prototype into an app that real users can call, with retrieval, tools and monitoring attached. Typical projects include:

• Internal knowledge assistants that answer from policies, wikis and shared drives
• Customer-facing chatbots embedded on a website with grounded answers
• Multi-step workflows that classify, extract and route documents or tickets
• Agents that call external tools and MCP servers to complete tasks
• Publishing a workflow as an API endpoint or MCP tool for other systems
• Comparing models and prompts using logs, feedback and annotations

Who Dify is best for

Dify suits developers, AI engineers and technical product teams who want a visual builder without giving up control of hosting. Companies with data residency rules often pick it because the Community Edition runs on Docker and the enterprise edition supports VPC and Kubernetes deployments with SSO and RBAC. Business users who just want an assistant to automate email and CRM tasks may find it more technical than they need, and teams committed to writing everything in code may prefer a framework.

Dify pricing

As of October 2026, Dify Cloud has a free Sandbox plan with 200 message credits, 5 apps, one team member and 50 knowledge documents. The Professional plan costs $590 per workspace per year and includes 5,000 message credits a month, 50 apps, 3 team members and 500 knowledge documents. The Team plan costs $1,590 per workspace per year with 10,000 monthly credits, 200 apps and up to 50 members. Monthly billing is also offered. After credits run out you can switch to your own model API key. The self-hosted Community Edition is free, and Enterprise is priced by quote with SSO, multiple workspaces and SLAs.

Dify limitations

• The open-source license is source-available with added conditions, so read it before building a commercial service on it
• Self-hosting means you own upgrades, backups and scaling
• Complex workflows can become hard to read on the canvas without naming and grouping discipline
• Cloud message credits are modest on lower tiers, so most production use needs your own API keys

Dify vs alternatives

n8n is the better pick when your main job is connecting business apps, with AI as one step among many integrations. LangChain suits engineers who want to write agents in Python or TypeScript and treat orchestration as code. LlamaIndex is stronger when document parsing and retrieval quality are the core problem. Relevance AI targets business teams building sales and operations agents with less setup. Dify wins when you want a visual LLM app builder, built-in RAG and the option to self-host it all.

Tips for getting started with Dify

Start on the cloud Sandbox to learn the editor, then move to the Community Edition on Docker once you know you need self-hosting. Build your knowledge base first and test retrieval on real questions before wiring it into an agent. Add your own model provider key early so credit limits do not interrupt testing. Use annotations on logged conversations to fix recurring bad answers.

Dify FAQ

Is Dify free?

Yes. Dify's Community Edition is free to self-host, and Dify Cloud has a free Sandbox plan with 200 message credits and 5 apps. Paid cloud plans start at $590 per workspace per year.

Is Dify open source?

Dify publishes its code publicly under a source-available license derived from Apache 2.0, with extra conditions, so check the terms before reselling it as a service.

Can I self-host Dify?

Yes. The Community Edition runs on your own servers with Docker, and Dify Enterprise adds VPC deployment, a Kubernetes Helm chart, SSO and dedicated support.

Dify vs LangChain: which is better?

Dify is better if you want a visual builder with RAG and hosting included, while LangChain is better for engineers who prefer to build agents entirely in code.