Dify
Open-source platform from LangGenius for building LLM applications in a visual interface: agent workflows on a canvas, RAG pipelines that ingest PDFs and slide decks, management of hundreds of models, a prompt editor, observability wired to Langfuse, Opik or Arize Phoenix, and an API to plug it all into your product. In the cloud on Dify Cloud, or self-hosted with Docker Compose (2 cores and 4 GB of RAM minimum). Dify Open Source License, based on Apache 2.0 with additional conditions.
Strengths
- Agent workflows, RAG, model management and observability together in one interface
- Every step of the flow is traced, which makes a bad answer explainable
- Cloud or self-hosted, with an API to plug apps into your product
- Model agnostic: hundreds of LLMs and any OpenAI-compatible model
Limitations
- Modified licence: no multi-tenant service or unbranded console without a commercial agreement
- Oversized for a simple chain of tools, where n8n stays more direct
- When self-hosting, one more server to run and keep updated
Best for
- An assistant that answers customers from your own knowledge base
- A PM who wants to prototype an agent flow and show it to the team without writing the orchestration