ACPpotamia

Features / AI governance and compliance

AI governance across every team that runs agents.

When several departments run AI agents, security and finance need one place to set rules and see what happened. ACPpotamia groups teams into an organization with one model policy, one kill switch, one view of spend and one compliance export, while each team keeps its own keys and limits.

How it works.

Approved models only

Allow or block models by pattern for the whole organization, each team and each API key. Refused calls never reach the provider.

Organization kill switch

Stop every agent in every team at once during an incident, and revive them with one click.

Central audit and export

Every call from every team, blocked ones included, exported as CSV, JSONL or raw daily archives, from the console or an API for your SIEM.

Roles and an audit trail

Viewers, admins and owners at organization level; every organization change is logged with who made it.

Safe testing

Test keys run every policy and limit with mock responses, so CI never calls or pays a model provider.

Questions.

Can we restrict which AI models employees use?

Yes. Set an allow list such as claude-* and gpt-5* for the whole organization; teams and keys can narrow it further but not widen it.

Can we export logs for compliance?

Yes. Download every call as CSV or JSONL, or pull it from GET /v1/controlplane/audit/export with an admin key. Records are kept for 365 days.

What happens to provider keys?

They are passed through on each call and never stored or logged.

Related features.