AnythingLLM is an open-source AI assistant that runs entirely on your own computer or server. Download the desktop app for macOS, Windows or Linux, pick a model or let it recommend one for your hardware, and start chatting with your documents without creating an account or adding an API key. It builds a private knowledge base from your files, scrapes and searches the web, runs custom agent skills and includes a Meeting Assistant that transcribes and summarizes calls locally. The desktop app and Docker self-hosting are free, while hosted cloud instances cost $50 per month for Basic and $99 per month for Pro.
People who care where their data goes get the most from it: researchers with confidential documents, consultants, lawyers, developers and small teams that cannot send material to a third-party cloud. Privacy-minded individuals who simply dislike token limits and subscriptions use it as a personal assistant. IT teams can self-host a multi-user deployment with isolated workspaces and admin controls, and companies that need white-labeling or SSO and RBAC can ask about the Enterprise package.
Common uses include loading contracts, reports or notes and asking questions about them, running a research agent that browses the web, and recording a Zoom or Teams meeting without a bot joining the call. Developers use the custom tool builder and developer API to wire AnythingLLM into their own workflows, and Windows users can run models on CPU, GPU or NPU through Microsoft Foundry Local. An Android app brings small on-device models, document chat and scheduled jobs to a phone, and it can pair with the desktop for larger models.
