Chat with documents and run extensible AI workflows on your own computer
AnythingLLM is a local-first AI workspace that brings models, documents and agent tools into one application. The desktop version runs on macOS, Windows and Linux, while a self-hosted deployment supports teams that want to operate the service on their own infrastructure. It is intended for people who want an assistant grounded in their files without making a hosted chat subscription the only way to work.
Users can build a document knowledge base and ask questions against their material. The application supports local model selection, including a recommendation based on the computer, and can also connect to external model providers. These choices affect both capability and privacy: a fully local configuration keeps inference on the device, whereas a selected cloud provider processes requests under its own service terms. Document retrieval should therefore be configured together with the model and embedding settings, not treated as an independent checkbox.
AnythingLLM includes web research, file-system tools, custom agent skills and scheduled jobs. Its documentation also covers MCP compatibility and integrations with email and calendar services. The meeting assistant can capture local meeting audio and produce transcripts, summaries and action items without placing a bot in the call. The Magic features extend assistance beyond a chat window with dictation, selected-text actions and context-aware completion. Availability and limits depend on the desktop version and feature tier.
The core desktop application does not require an account or cloud API key to begin using local models. The self-hosted project is MIT licensed, while optional Desktop Pro removes daily limits from the Magic features and adds related benefits. Hosted deployments are another option for teams that do not want to manage infrastructure. Local performance depends on available hardware and the chosen model, so a smaller model and a representative document collection are useful starting points before expanding an installation.
