// the find
Mintplex-Labs/anything-llm
Stop renting your intelligence. Own it with AnythingLLM. Everything you need for a powerful local-first agent experience
AnythingLLM is a self-hosted RAG and agent platform — essentially a private ChatGPT you run yourself, with document ingestion, multi-user support, and a no-code agent builder. It targets teams and individuals who want to run chat-over-docs workflows without sending data to a third party. The scope is unusually wide: 30+ LLM providers, 10 vector databases, desktop app, browser extension, mobile app, and an embeddable widget.
The provider abstraction is genuinely useful — swapping from OpenAI to Ollama to Bedrock is a settings change, not a code change. The separation of a document collector service from the main server is a sound architectural call that keeps parsing failures isolated. MCP compatibility means it plugs into the broader tool ecosystem without requiring custom integration work. The intelligent tool selection feature (claiming 80% token reduction by not passing all tools to every query) addresses a real and annoying problem in agentic workflows.
The monorepo spans three separate Node services plus submodules, and the dev setup requires running them all in parallel — the barrier to contributing or debugging anything non-trivial is high. JavaScript throughout means the RAG pipeline has no real type safety; the document parsing code in particular looks like it accumulated connectors faster than it accumulated tests. Multi-user permissioning is Docker-only, which is an odd constraint that suggests the feature was bolted on rather than designed in. With 30+ LLM integrations, the maintenance surface is enormous and individual provider support tends to lag when APIs change.