Summary
One process table shared by your agents and your terminal.
An agent-launched process normally dies with the tool call and its output goes with it; mcproc keeps it running, persists the logs, and lets both the agent and you reach it afterwards. Wait-for-log is the detail that saves the most time — the start call returns when the readiness pattern appears, not after a fixed sleep.
What it is
A Rust server and CLI pair for background processes: an agent starts and stops long-running processes through MCP, and you inspect and control those same processes from the shell.
What you get
- Processes started, stopped and restarted as named background jobs
- Every process listed with its status, visible to the CLI and to other agents alike
- Logs captured and persisted, then read back or searched with regex, with time filtering and context lines
- Wait-for-log at start, so a process is reported ready only once the pattern you named appears
- Processes grouped automatically by project context
- Commands executed through version managers — mise, asdf, nvm, rbenv
- Child processes cleaned up automatically when the parent is stopped
Requirements
`brew tap neptaco/tap && brew install mcproc`, or build from source with the Rust toolchain and a protobuf compiler. Files follow the XDG Base Directory specification.
Setup effort
One command — brew install mcproc