Attach, don't migrate
Point AgentPod at agent runtimes you already run. It detects them in place — Claude Code, Codex, OpenCode, Hermes, OpenClaw and Pi — rather than asking you to move anything.
The console for the machines your agents live on
Your agents run somewhere. AgentPod is how you see that place, and reach into it.
Reaches machines you can't
Every node dials out to the hub over a WebSocket. No inbound ports, no port forwarding, and laptops behind NAT or CGNAT work the same as a VPS.
Files, logs, terminal, health
Read a workspace, tail a log, open a shell, check what’s running, plan a cleanup — on any station in the fleet, from one place.
Talk to them in a room
Turn on the bridge and every station gets a chat identity of its own. Message it from your phone, send it a voice note, answer its permission questions with a number — and get the provider’s own sentence when a turn fails. The bridge
Skills you can prove loaded
Install a skill under review, then verify a fresh session actually loaded it. Present, eligible and loaded are separate observations, never inferred from one another. Managed skills
Hosting requirements
The hub is a Bun service with a Postgres database, and the console is a static site. Current releases also require configuration and a credential from an external account service that is not public source. Start with a hub you already have access to.
These docs are also published as plain text for agents: /llms.txt indexes them,
and /llms-full.txt is every page in one file.