Skip to main content
Early Access. Commands and behavior can change between releases.
Monitor is a tool, not a command: the model uses it when you ask it to watch something, such as a dev server, a build, a log file or a flaky test loop. The command runs in the background, and each stdout line arrives as a notification while the model keeps working. So “start the server and tell me when it errors” is one instruction.
A watch ends when the process exits, when it times out (5 minutes by default, up to an hour), when the agent cancels it, or when stdout floods (more than 200 lines in 10 seconds). For a log tail or anything else that should keep going, ask for a watch that lasts the whole session. It has no timeout, and still ends when the process exits, floods or the session closes. Ask the agent what it’s watching: it can list, inspect or cancel running watches. Inspecting a watch shows how many lines it has sent, so you can spot a missing notification. Monitor shows in the transcript like any built-in tool, and sub-agents can use it too.
Monitor watches processes. It is not the orchestrator’s /monitor command, which watches deployments. To watch a process, just ask in plain words.

​
Include errors written to stderr

Only stdout lines become notifications and count toward the flood limit. Stderr is logged, not sent. If the command writes errors to stderr and you want the agent to react to them, redirect stderr into stdout when you start the watch, for example bun run dev 2>&1.

​
Inspect a finished watch

The agent uses Jobs to list, inspect or cancel watches. The last 20 finished watches keep their exit details, so a result is still there after its notification scrolls past. Notifications that arrive while the model is busy are grouped and delivered together next. Monitor works in normal Architech and agent sessions, including sub-agents. Prime has no Monitor or Jobs; it uses its own code-based process tools.