Commands
Run these in your own terminal. The ones that change what agents may do need a terminal, and the default policy denies them to agents.
Everyday
Section titled “Everyday”| Command | What it does |
|---|---|
moat |
Which agents are protected, today’s decisions, and anything that needs you |
moat allow --last |
Allow the last ask for the rest of that agent session (24 h) |
moat allow --last --always |
Allow it for good: a rule for exactly that command or those files |
moat allow --dir <folder> |
Read and write a folder outside the project |
moat allow --site <host> |
Network access to one host |
moat allow --remove <id> |
Take back a rule moat allow added |
moat secret add gh |
Keep the GitHub token from the agent |
moat policy check "<command>" |
The verdict for a command, without running it |
See what happened
Section titled “See what happened”| Command | What it does |
|---|---|
moat show [id] [--since today] |
Recent events, or one in full |
moat replay --since today |
Each agent session as a timeline |
moat report --since 7d |
Totals, hosts, top rules, asks per hour |
moat audit export · moat audit verify <file> |
Export events with their chain hashes · check an export |
Set up and maintain
Section titled “Set up and maintain”| Command | What it does |
|---|---|
moat init |
Install the policy, lock and log; asks before hooking each agent |
moat status · moat doctor |
Each agent’s protection level · what drifted, and re-pin with --accept |
moat edit |
Edit the policy; checked and diffed before it applies |
moat policy lint |
Check a policy |
moat policy upgrade |
Take what a new release’s default adds, keeping your edits |
moat sandbox show · moat sandbox sync |
The sandbox settings the policy compiles to · write them for your agents |
moat trust [--revoke] |
Let a repository’s .moat/policy.yaml allow |
moat uninstall [--hosts …] [--purge] |
Remove OpenMoat and restore every settings file |
Advanced
Section titled “Advanced”| Command | What it does |
|---|---|
moat run [--isolate] -- <agent> |
Run a whole agent in a sandbox generated from the policy |
moat proxy · moat proxy install |
The egress proxy, on its own or as a user service |
moat bench |
Score OpenMoat, or any hook, against the bundled attack scenarios |
moat guard --host <id> |
The hook entry point: agents call it, you do not |
Every option: USAGE.md.