langgenius/dify
Production-ready platform for agentic workflow development.
Dify is an open-source LLM app platform by LangGenius that combines visual workflows, RAG pipelines, agents, and LLMOps in one self-hostable canvas. If you want production AI apps without writing glue code for every model integration, start here.
Our Review
Dify is the most-starred open-source LLM application platform on GitHub -- 141,000 stars and 22,000 forks as of May 2026. Built in TypeScript by LangGenius (a Menlo Park company that raised $30M Series Pre-A at a $180M valuation), it gives you a visual canvas to design AI workflows, RAG pipelines, and agents, then ship them as APIs without writing the glue code.
What Dify does:
- •Visual workflow canvas drag-and-drop nodes for prompts, model calls, RAG retrieval, conditional branching, and tool calls.
- •Hundreds of LLM providers OpenAI, Anthropic, Google, Mistral, Llama, DeepSeek, Qwen, and any OpenAI-compatible endpoint including self-hosted Ollama and vLLM.
- •Built-in RAG pipeline document ingestion for PDFs, PPTs, Word, Markdown plus extraction, chunking, vector storage, and reranking, all configurable from the UI.
- •Agent framework with 50+ tools ReAct and function-calling agents with pre-built integrations for Google Search, DALL-E, Stable Diffusion, WolframAlpha, plus custom tool support.
- •Prompt IDE side-by-side prompt testing across models with version history, useful for debugging hallucinations before production.
- •LLMOps observability application logs, latency tracking, cost monitoring, and native integrations with Opik, Langfuse, and Arize Phoenix.
- •Backend-as-a-Service every Dify app exposes a clean REST API so existing frontends or backends can call it without rewriting business logic.
Dify ecosystem:
- •Model providers OpenAI, Anthropic, Google, Mistral, plus self-hosted via Ollama, vLLM, and any OpenAI-compatible API.
- •Observability Opik, Langfuse, and Arize Phoenix for trace and metric forwarding.
- •AWS Marketplace one-click deploy to your own VPC via Dify Premium AMI for startups and small businesses.
- •Self-hosted Helm charts community-contributed Kubernetes deployment for HA setups.
Getting started:
Install with Docker Compose: cd dify/docker && cp .env.example .env && docker compose up -d, then open http://localhost/install. Minimum requirements are 2 CPU cores and 4 GB RAM. Full docs at docs.dify.ai. Cloud version at dify.ai gives 200 free GPT-4 calls on the sandbox plan if you want to try without installing.
Limitations:
The license is a modified Apache 2.0 with a multi-tenant SaaS restriction -- you cannot offer Dify as a hosted service to your own customers without commercial agreement. Self-hosting requires Docker and basic ops knowledge -- not appropriate for non-technical users. The agent framework is good but less flexible than code-first frameworks like CrewAI or LangGraph for complex multi-agent orchestration. Some enterprise features such as SSO, audit logs, and multi-workspace governance sit behind the paid Cloud or Enterprise tiers.
Our Verdict
Dify in 2026 is the default open-source choice for teams shipping production LLM applications without rebuilding the orchestration layer from scratch. 141,000 GitHub stars, 22,000 forks, weekly releases, and a $30M Series Pre-A round at a $180M valuation make this one of the most active platforms in the space.
For Vibe Builders, Dify removes the need to code agent loops or wire up RAG by hand. The visual canvas, pre-built tool integrations, and Docker-based quickstart get you from idea to working AI app in an afternoon. Pair it with Ollama for fully local inference and you have a private AI app builder running on your own machine.
For Developers, the value is the platform abstractions plus the escape hatches: every app exposes a REST API, custom tools plug in cleanly, and you can self-host the whole stack with Docker Compose or Helm. The LLMOps layer (Opik, Langfuse, Arize Phoenix integrations) covers production observability without bolting on a separate tool.
Skip Dify if you need fine-grained code control over agent execution -- LangGraph or CrewAI give you that and Dify intentionally does not. Skip if you need a managed SaaS where you resell Dify as a service to customers -- the license forbids it. Skip if your team has zero Docker experience and you want a desktop app, in which case AnythingLLM or LM Studio is a better fit.
Frequently Asked Questions
What is Dify?
Dify is an open-source LLM application development platform built by LangGenius. It combines a visual workflow canvas, RAG pipeline, agent framework with 50+ built-in tools, and LLMOps observability in a single self-hostable system. As of May 2026 it has 141,000 GitHub stars and ships weekly releases.
Is Dify free and open source?
Dify Community Edition is open source under a modified Apache 2.0 license. The modification adds a multi-tenant SaaS clause that prevents you from reselling Dify as a hosted service without a commercial agreement. Dify Cloud and Dify Enterprise are paid tiers with additional features like SSO and audit logs.
Dify vs Langflow vs n8n -- which should I choose?
Choose Dify when you want a visual no-code platform built specifically for LLM apps with RAG, agents, and observability included. Choose Langflow when you need LangChain-native flows for complex multi-agent and vector pipeline work. Choose n8n when your primary need is general workflow automation across 400+ SaaS tools with LLM enrichment steps mixed in.
How do I install Dify?
Install Dify locally with Docker Compose. Clone the repository, navigate to the docker directory, copy the .env.example file, and run docker compose up -d. The minimum requirements are 2 CPU cores and 4 GB of RAM. Open localhost in your browser to start the setup wizard. Full documentation lives at docs.dify.ai.
Who built Dify and where is the company based?
Dify is built by LangGenius, a startup founded by Luyu Zhang. The company is now headquartered in Menlo Park, California while its core open-source engineering team of around 60 remains in China. LangGenius raised $30 million in Series Pre-A funding at a $180 million valuation.
What license does dify use?
dify uses the Apache-2.0-modified license.
What are alternatives to dify?
Explore related tools and alternatives on My AI Guide.
Open source & community-verified
Apache-2.0-modified licensed: free to use in any project, no strings attached. 142,043 developers have starred this, meaning the community has reviewed and trusted it.
Reviewed by My AI Guide for relevance, quality, and active maintenance before listing.
Topics