AnythingLLM

Open SourceMITSelf-hostable

The all-in-one AI workspace — chat with your documents, build agents and run private RAG with any model, commercial or local. A self-hosted alternative to hosted AI assistants.

Open source alternative to

Claude

ChatGPT

Visit AnythingLLM View source

About AnythingLLM

AnythingLLM turns 'chat with your documents' from a SaaS subscription into software you own. Drop in PDFs, docs and websites, pick any model — GPT, Claude, Gemini or a local Ollama model — and get accurate, cited answers over your own knowledge.

It's a full AI workspace: multi-user support with permissions, AI agents that can browse and run tools, and a no-code agent builder. Run it as a desktop app or self-host the server; with local models, nothing ever leaves your machine.

MIT-licensed with one of the largest communities in local AI, AnythingLLM is the practical way for teams and tinkerers to get private, document-grounded AI.

Key features

Chat with anything — PDFs, docs, websites and more become a knowledge base your AI answers from, with citations.
Any model, your choice — OpenAI, Anthropic, Gemini or fully local models via Ollama — swap anytime.
Agents & tools — Built-in AI agents that browse, summarize and act, plus a no-code agent builder.
Private & multi-user — Self-host with workspaces and permissions; local models keep data fully on-device.

The honest read

What we like

Genuine private RAG over your documents
Works with commercial and local models
Multi-user with permissions
Desktop app or Docker server
MIT licensed, huge community

Worth knowing

Answer quality depends on the model you pick
Local models need decent hardware
UI is functional over fancy
Advanced RAG tuning takes reading

FAQ

Is AnythingLLM free?

Yes — MIT-licensed and free whether you use the desktop app or self-host the server. Your only cost is model usage, or nothing with local models.

Can it run fully offline?

Yes — pair it with Ollama and a local model, and documents, embeddings and chats never leave your machine.

How is it different from ChatGPT?

AnythingLLM is open source and grounds answers in your own documents with citations, using whichever model you choose — instead of a closed assistant with your data on someone else's servers.

Does it support teams?

Yes — multi-user workspaces with permissions make it usable as a shared internal AI over company knowledge.

What are its agents?

Configurable AI agents that can search the web, scrape pages, summarize and chain tools — buildable without code.