with one click
lopi
// Autonomously manage instances via the lopi CLI — env, status, troubleshoot. Use when the user asks to debug, operate, or inspect their system.
// Autonomously manage instances via the lopi CLI — env, status, troubleshoot. Use when the user asks to debug, operate, or inspect their system.
| name | lopi |
| description | Autonomously manage instances via the lopi CLI — env, status, troubleshoot. Use when the user asks to debug, operate, or inspect their system. |
You are operating a system managed by the lopi CLI. The project's agent context lives in tools/lopi/AGENTS.md. Your job is to triage and act with well-calibrated caution.
Ask the user unless they've already told you. Three modes:
lopi <subcommand> directly from the primary repo.tools/lopi/lopido <instance> <subcommand> --json — resolves the instance via scripts/resolve.sh, SSHes in, runs lopi there.lopido fails with a resolution error, stop and ask the user to check scripts/resolve.sh.Always start by running:
lopi command-tree --json [--read-only]
This is the source of truth for what's available. In read-only mode, pass --read-only to filter out write commands. Stub commands that aren't yet wired up for this system return {"success": false, "error": "not-configured", ...} — surface the hint, don't retry.
--json for structured output; parse it.lopi intel search <keywords> --json — if you find a hit, apply the workaround.lopi exec -- <cmd> runs an arbitrary command from the project root. Useful for shelling out to runtime tools (docker compose ps, systemctl status <unit>) when lopi's wrappers are still stubs.
lopi status --json (if configured, else lopi exec -- docker compose ps or systemctl status <unit>)lopi troubleshoot all --json (if configured)lopi intel search <keywords> --json