Talk to an LLM that can run tools on your devices. Register a laptop, a workstation, a home server, a Raspberry Pi — install a small Go agent and the model gets controlled access to the local filesystem and shell. Automate the boring parts of running a homelab: nightly backups, container restarts, log triage, sensor reads. Save useful instructions as workflows; trigger them by hand or on a cron schedule.
Stream conversations with Claude. When you ask about something local — a file, a directory listing, the host OS — the model calls the right tool on the right device and answers from the live data.
Save an instruction with a name. Run it on demand, or attach a cron expression and let it fire on a schedule. Every run is recorded; output is one click away in the history.
Install a small Go binary on any machine. It registers with the server over HTTP and advertises its tools. The server brokers every call; the agent only runs what's allow-listed.