AnythingLLM
Mintplex-Labs · AI assistants
GitHub stars
67,713Awaiting first sync
No measured star history yet
What is AnythingLLM?
AnythingLLM gives you a ChatGPT-style assistant that you run yourself and point at your own documents. You choose the model, from a local one on your own computer to a cloud provider, and keep files on hardware you control. One person can start with the desktop app, while a team can run the Docker server with separate users and permissions.
AnythingLLM is an all-in-one AI application from Mintplex Labs for chatting with documents and running AI agents. It runs local models through its built-in engine or Ollama, or connects to cloud providers such as OpenAI and Anthropic. It comes as a desktop app for macOS, Windows and Linux, or as a Docker server with multiple users and roles.
Core capabilities
- Chat with PDFs, Word files and text documents, with source citations in each answer.
- Build agents with a no-code flow editor, custom skills and MCP servers.
- The Docker version adds user roles, white-labelling and an embeddable website chat widget.
Documents, models and workspaces
Documents are added to workspaces by drag and drop, then embedded into a vector database, with LanceDB as the default and options such as PGVector, Pinecone, Chroma, Qdrant and Milvus. Chat answers show citations back to the source files. The README lists many model providers, including OpenAI, Azure OpenAI, AWS Bedrock, Anthropic, Google Gemini, Mistral, Ollama and LM Studio. The Docker image can also run local models itself on CPU or GPU through an optional built-in engine.
- RAG
- Vector database
- Citations
- Local models
Agents, flows and scheduled tasks
Agents inside a workspace can browse the web and use custom skills. Agent Flows is a visual, no-code editor with blocks such as Web Scraper, API Call, LLM Instruction, Read File and Write File. A flow runs when you use the @agent directive or ask a related question in an agent chat. AnythingLLM can also use tools from MCP servers, run recurring prompts on a cron schedule and be reached through a developer API.
- AI agents
- Agent Flows
- MCP
- Scheduled tasks
- API
Desktop app or Docker server
AnythingLLM is licensed under MIT. The desktop app installs on macOS, Windows and Linux for a single user, while the mintplexlabs/anythingllm Docker image serves the web app on port 3001 and keeps its data in a mounted storage folder. Multi-user mode, white-labelling and the website chat widget are only in the Docker version. Mintplex Labs also sells AnythingLLM Cloud, a paid hosted instance it manages for you. The app sends anonymous usage telemetry by default, and the README explains how to opt out.
- MIT
- Docker
- Desktop app
- Self-hosted
Where it fits
Use cases
- 01
Answer staff questions from company documents
Create a workspace, then drag and drop PDFs, DOCX or text files into it so AnythingLLM embeds them in its vector database. Ask questions in the workspace chat and check the source citations under each answer. Pick a local model through the built-in engine or Ollama to keep the data on your own hardware.
- 02
Give each team access to only its own assistant
Run the Docker version and turn on multi-user mode, which creates the first admin account. Add people as Admin, Manager or Default users; Default users can only chat in the workspaces they have been added to. Multi-user mode cannot be switched back to single-user mode, so plan this before you start.
- 03
Add a document-based chatbot to your website
In the Docker version, create an embedded chat widget based on a workspace and choose Query mode so it only answers from that workspace's documents. Set the allowed domains and the maximum chats per day or per session. Then paste the script tag AnythingLLM gives you into your site's HTML.