Why you need an Agentic Control Plane
Your agents stopped chatting and started acting — refunds, database writes, emails to customers. The checks you hand-rolled hold only while you're watching. A control plane makes them hold when you're not: every call identified, checked against your rules, metered, and logged.
What it looks like when this bites
This one is ours. From Caught in the Wild, the running log of real interceptions on our own agents:
The surprise bill came from the one agent outside the plane
“A ~$15/day model bill traced to a container running on a fresh API key — spend invisible because it never crossed the proxy. … The bill you get surprised by is always from the agent you didn't control.”
Nothing crashed. No alert fired. The agent did its job — on a key nobody was metering, at a run rate nobody chose. Every check we'd written held; the failure was the call path those checks never saw. That's the shape of the problem: agents don't fail loudly, they act quietly, and whatever isn't on the call path is invisible until the invoice, the deleted row, or the leaked record arrives.
Read the full log →Before and after
| Concern | Without an ACP | With an ACP |
|---|---|---|
| Identity | Shared API key — backend can't distinguish users | Every request bound to verified user identity |
| Authorization | All-or-nothing access, enforced in app code | Per-user, per-tool policies enforced at the gateway |
| Data protection | PII flows into prompts unchecked | PII detected and redacted before reaching the model |
| Cost control | No per-user limits — surprise bills, runaway loops | Rate limits, budget caps, and agent guard per user |
| Audit | Generic logs with no user attribution | Every action logged with identity, policy, and cost |
| Compliance | Manual controls, audit gaps, failed reviews | Automated evidence for HIPAA, SOC2, GDPR, PCI |
Thirty seconds to the first audit row
One command, no account, on your machine — or self-host the MIT-licensed modules. Same control engine, your choice of deployment.