Dify

Dify is an open-source LLMOps and AI-app development platform that unifies a visual workflow builder, RAG pipeline engine, agent framework, prompt IDE, and observability. It supports hundreds of models across many providers and can be self-hosted via Docker or used through a hosted cloud service, making it popular with teams that want to ship LLM apps quickly while keeping control over their data.

Dify positions itself as a 'Backend-as-a-Service plus LLMOps' layer for AI development. Rather than stitching together separate tools for prompting, retrieval, orchestration, and monitoring, it bundles them into one workspace with a low-code visual builder. Teams use it to build chatbots, agents, and multi-step workflows that call hundreds of models from dozens of providers, including self-hosted open models. The project is one of the most-starred open-source AI repositories on GitHub and reportedly runs across more than a million machines worldwide. It ships a Community Edition that is free to self-host via Docker, plus a hosted cloud service and paid Enterprise tiers that add SSO, multi-tenant management, and provider load balancing. Dify's license is a modified Apache 2.0: you can use and self-host it commercially for your own apps, but you may not run it as a multi-tenant resale service or remove its branding without authorization. In practice the platform fee is often a small fraction of total cost, since the underlying LLM token spend bills separately through your own provider keys.