Clawgate checks every request, applies your rules, and records the cost before a single token is billed. Here is everything it does.
Limits are checked before each request runs, so spend simply cannot go over what you set. Cap usage per person, and set an all-time spend ceiling per project.
When a cap is hit, the request is blocked with a clear message, not a silent retry. Owners are never surprised by the cost.
If a request asks for a model you don’t allow, Clawgate quietly serves the best allowed model instead of failing. Your team keeps working, and you keep control of cost.
Claude Code, Codex, and OpenCode pick a model on their own, and that choice drives your cost. Clawgate puts the decision back in your hands, so you can serve lower-cost models when full power isn’t needed.
Your coding agent fires a lot of routine requests that don’t need your most expensive model. Turn on smart routing and Clawgate reads each request, judges how demanding it is, and serves the best allowed model for it, so the easy work runs on a lighter, cheaper model and your top model is saved for the hard parts. One switch, no tuning.
Routing only ever picks from the models you allow, so you keep control while Clawgate quietly serves the cheapest model that fits each request.
Every request is tied to a project and a user. No more one big number on the invoice. See exactly where the spend goes, and stop paying for side projects.
x-project-idCost is tracked to the micro-dollar for each model, including separate cache rates, so the numbers are exact, not estimated.
Detailed events follow your plan’s retention (7 days to unlimited). Monthly summaries are kept for good, so your invoices and history survive even on shorter plans.
Track tokens, cost, latency, and sessions as they happen. Export to CSV (Team+) for your own analysis and chargeback.
Clawgate fingerprints request content with privacy-safe hashes, never raw code, to flag suspicious patterns for your review.
Signals are scored low, medium, or high and sent to a review queue. They are never enforced automatically. You decide what to do. Scan frequency scales with your plan, up to nightly.
Clawgate can compress bulky tool output before it reaches the model. Old context gets squeezed, fresh work stays untouched, and accuracy holds. One switch per organization, with savings shown on your dashboard in dollars.
Tokens per request on real agent workloads:
| Workload | Before | After | Savings |
|---|---|---|---|
| Code search (100 results) | 17,765 | 1,408 | 92% |
| SRE incident debugging | 65,694 | 5,118 | 92% |
| GitHub issue triage | 54,174 | 14,761 | 73% |
| Codebase exploration | 78,502 | 41,254 | 47% |
Platform fees and per-seat fees are shown clearly at billing time. The AI cost stays pure pass-through, cache savings reach you, and failed requests are never billed.
Monthly invoice snapshots with per-model and per-user breakdowns. Control who can do what with roles and permissions.