Agents
Start a coding agent on your box, close the laptop, and let it work; it is an ordinary process on a machine you already pay for, so nothing it does is metered.
On this page9
Start an agent
prized agents run claudeThis starts Claude Code in a detached tmux session named claude, in your synced project, and prints the reattach command; closing the lid changes nothing. A sleeping or paused box wakes first. run never starts a second copy of a session that exists; it says so.
| Flag | What it does |
|---|---|
--attach | Land in the session as it starts; otherwise reattach later with prized ssh -s claude. |
--dir | Working directory. Defaults to your only synced project; a clone is not one, so pass --dir ~/code/<name>. |
--session | Session name, for a second concurrent run of the same agent. |
-- ARGS | Everything after -- goes to the agent. |
The agent's own copy action (Claude Code: select, then c) lands on your clipboard through prized ssh (Copy out).
The agents
| Agent | On the box | Runs on the box |
|---|---|---|
Claude Code (claude) | Preinstalled. | Yes |
Codex CLI (codex) | Preinstalled. | Yes |
OpenCode (opencode) | Installed on first use, from a pinned release. | Yes |
pi (pi) | Installed on first use, from a pinned release. | Yes |
Cursor (cursor) | Installed on demand (--install-cursor). | Yes |
ChatGPT app (chatgpt) | Drives codex over SSH from your machine. | No |
chatgpt is a connection, not a process on the box, so its tasks stop when your machine sleeps and prized agents run chatgpt refuses; prized agents run codex survives a closed laptop. App setup is on Codex remote.
OpenCode and pi
prized agents run opencodeThe first prized agents run or prized prompt of either installs a pinned release on the box, checked against its SHA-256 before anything is unpacked; a mismatch installs nothing and the command says why. A copy you installed on the box yourself is used instead. Neither has a handoff: sign in on the box (opencode auth login; or pi, then /login), or set a provider key as a Secret (Keys for other agents and backends). OpenCode also runs its free models with no key.
See what is set up
prized agents lsOne row per agent: installed and signed in on your machine and on the box, MCP servers found, and where it is running. For OpenCode and pi, signed in on the box means a sign-in or a provider key there. Read-only; takes --json.
Hand credentials to the box
prized agents handoff claude codexLists what would travel, asks, then copies the named agents' sign-in and MCP config to the box; prized setup runs the same step. Re-run it after changing MCP config. --gh adds your GitHub CLI token and git identity, --install-cursor installs Cursor's CLI first; what travels per tool is on Credentials.
Drive it remotely
prized prompt --provider claude "fix the failing tests and run them again"prized prompt runs Claude Code, Codex, OpenCode or pi headless on the box and streams its events until the verdict; prized events follows a run and prized interrupt stops it, also over HTTP (Prompt an agent remotely). A model written amazon-bedrock/…, openrouter/… or deepseek/… runs it on that backend (Models and backends).
Several agents at once
Each agent gets one session by default (claude, codex, opencode, pi, cursor), each in its own project via --dir. For two runs of the same agent, name the sessions:
prized agents run claude --session claude-api --dir ~/code/api
prized agents run claude --session claude-web --dir ~/code/web
prized ssh --lsBring your own agent
Any CLI agent runs on the box. Pick other in setup's agent step and it prints these steps with your box's name filled in:
prized sshopens the box, waking it if needed.- Install the agent's CLI there.
- Sign in there, once per box; the credential stays on the box disk and Prized never sees it.
- Run it detached in a named session:
prized ssh -s myagent, ortmux new-session -d -s myagent 'the-agent'from any box shell. Reattach withprized ssh -s myagent.
For an API key, prefer a Connector: the box gets a placeholder and the real value never touches its disk. The handoff and prized agents run do not cover your own agent; sessions work the same.
Review from your machine
In a synced project, the agent's edits stream into the folder on your machine as it works, so you review the diff in your own editor. The box wins sync conflicts, so a local edit that races the agent yields to its work; a clone has no mirror, so review it over git.