Integrations · AI coding agents
Your policy pack writes Claude Code managed settings and hooks, so every session follows your rules, and its changes still pass the same gates as a person’s.
Control levels
1Observe
We watch, record and flag.
Every tool call an agent makes through the Aspian gateway, logged against the human it works for.
2Gate
The tool waits for an approval.
Managed hooks ask Aspian before risky commands: allow, block or ask. Its changes wait for a human approval.
3Hold
Only Aspian holds the credentials.
Agents never hold production credentials. Production writes go only through approved plans.
Set it up
- 1Generate managed settings from your policy pack.
- 2Distribute them by MDM or commit them to the repo.
- 3Connect agents to the Aspian MCP gateway with an agent key tied to a named person.
Limits worth knowing
- Local hooks can be bypassed on a machine you don’t manage. The server-side gates still apply, and such commits are marked “agent, unmanaged device”.
More ai coding agents
See what shipped without approval in your last 90 days.
Install the GitHub App and the scan arrives in minutes: every repo scored, every unreviewed merge and unapproved deploy named. Nothing is blocked until you choose.
Ship like a startup. Prove it like a bank.