What the Control Plane does
The AI Control Plane is where AI operations become observable and controllable. Every interaction that flows through the AI Gateway lands here as structured usage — and you decide how it’s aggregated, who can see it, how long it’s kept, and what triggers a review.
Features
| Feature | What it gives you |
|---|---|
| Usage analytics | Volumes, success rates, latency, and trends per service, tenant, or feature |
| Cost tracking | Provider cost and credit consumption, broken down by the dimensions you care about |
| Audit logs | Structured records of requests, changes, and operational events |
| Tenant controls | Per-tenant limits, retention, and access settings |
| User limits | Caps and quotas scoped to individual users |
| Service limits | Limits scoped to services and features |
| Safety events | Records flagged for review by safety checks or policy violations |
| Data retention | Configurable retention windows for usage and metadata |
| Access control | Roles for who can view logs, edit policies, or trigger operational actions |
Why teams pick the Control Plane
- One view of AI operations. No more stitching together provider dashboards.
- Costs under control. See where spend is going and act before it surprises you.
- Designed for auditability. Capture the records reviewers and stakeholders ask for.
- Configurable controls. Turn on what your use case needs; leave the rest off.
How it fits
The Control Plane reads from the same data model as the AI Gateway and PromptOps — so every log line, cost entry, and audit record is tied back to a specific interaction, prompt version, and caller. Nothing is inferred after the fact.
Get started
Start building with the Control Plane, or read the developer docs to wire up observability for your services.