AnythingLLM

Every workspace holds its own document set, and AnythingLLM parses PDFs, Office files, spreadsheets, code, audio, and URLs into a searchable index. Answers come back with inline citations that point at the chunk they came from, so verifying a claim takes a click instead of a hunt. This is the core of the product: chat with documents offline, on your own disk, with no upload step.

The desktop app ships with an engine built on llama.cpp, managed through AMD's Lemonade, and it replaced the older Ollama dependency in late 2026 with a smaller download and better Windows support. On first run it checks your RAM and GPU and recommends a model that will actually load. You can accept that, download a GGUF from Hugging Face, or connect a cloud provider. That flexibility is the reason the app stays useful as your hardware changes.

Calls get recorded, transcribed, and summarized on your machine, with speaker labels and a list of action items at the end. It works with any meeting platform because it listens to system audio rather than joining as a participant. That distinction matters: nobody sees an unfamiliar bot in the attendee list, and the recording never lands on a vendor's server. Transcripts get indexed into your workspace, so you can ask about a call from three weeks ago.

Agents can search the web, read workspace documents, run code, and write files. You build custom tools in plain JavaScript, and background jobs run prompts on a schedule, which covers recurring work like a morning digest of new documents. The AnythingLLM API exposes the same capabilities over HTTP for the Docker build, so you can trigger a workspace from a script or another app.

The Docker image turns the single-user app into a shared server. Each user gets isolated workspaces, admins control which models and tools are visible, and the whole interface can carry your branding. Enterprise deployments add SSO and role-based access control. For a company that wants a private AI workspace without building one, this is the shortest path.

AnythingLLM Mobile runs small GGUF models directly on Android phones, with web search, document retrieval, and scheduled jobs available offline. Pair it with your desktop over the local network and the phone reaches your full workspaces and larger models. Chats, files, and threads stay in sync without an account or a third-party relay.