Skip to content

fleet documentation

fleet lets your coding agent use every machine you have. You install it once, on the machine you work from, and then talk to your agent; these pages are for when you want to know more.

New here? Start with Getting started: one machine added, and a first job on it, in about ten minutes.

Guides

Task by task: what to say to your agent, and the command it runs.

Guide For when you want to…
Adding machines add a server, a rental, one from your ssh config or tailnet; names, tags, clones, removing
Reaching your machines let fleet in: Tailscale SSH, Windows and Mac, rentals, LAN; what to do when a machine refuses
Finding the right machine let the agent pick by GPU, memory or OS; tag machines; track costs and idle rentals
What is on each machine see GPU jobs, served models, installed software, and where models and data live
Running work run commands and long jobs on another machine
Agents add fleet to a specific agent or app — 35+ of them
What fleet changes see exactly what fleet writes on each machine, and take it back
Installing and upgrading install by hand, upgrade, put fleet on more machines, remove it
Upgrading from 0.5 what changed in 0.6, and removing what 0.5 left on your machines
Windows watch Windows machines, or run fleet on Windows
Troubleshooting find out what a message means and what to do

Reference

Exact and complete; generated from the code where it can be.

Design

Why fleet works the way it does.

  • How fleet works — what happens, step by step, when you install, add a machine, measure, diagnose a refusal and run a job; and which code does it.
  • Core, skill and MCP — one fleet per machine, shared by every agent.

Contributing

CONTRIBUTING.md covers running the tests, the end-to-end checks on each OS, and the files that are generated rather than written.