Giving coding agents access to your Mac, with limits you control
Coding agents understand your repository well. They know very little about the machine it runs on.
Ask one to free up disk space and it often reaches for rm -rf. Ask what is using port 3000 and it pipes raw lsof output into the conversation.
Room Service 3.3 includes a local MCP server that gives Claude Code, Codex, and Cursor a safer way in.
Agents can now see:
what is taking space, with risk labels
which dev servers and watchers are running
project storage and Git state
They can also act: stop a port, quit a stuck build, or move build data to the Trash. Every action runs through the same code as the app's own buttons.
The limits are built in:
Off by default. Nothing listens until you enable it.
No network exposure. It is an owner-only local socket.
Permanent actions are disabled by default. When enabled, each one requires approval in Room Service.
Developer Vault is never exposed. No tool can read it.
Every agent action is recorded in the Ops Feed.
How much access do you currently give your coding agents?
Replies
Love that the Ops Feed logs every agent action.. Thanks for building this with real limits.
Does stop a port recheck which process owns it before acting?
A dev server can restart between inspection and the stop request so I would want a changed process to require a fresh decision
@al_kub Good catch, and partly yes today.
After sending SIGTERM, Room Service rescans the port. If a different PID is listening by then, it stops there and reports it instead of touching the new process. For agents it also never escalates to a force quit.
What it doesn't do yet is confirm the owner before the first signal. stop_port takes only a port number, so if the server restarted after list_ports, the new process could get the SIGTERM.
I'm changing that: stop_port will take the PID the agent saw, rescan right before acting, and refuse with "owner changed, check again" if they don't match. Thanks for pointing it out 🙏