GitHub Copilot Credit Tracker
Track GitHub Copilot AI Credits before they disappear.
Premium-request credits vanish fast once your team starts using GPT-4.1, Claude, or Gemini through Copilot. AgentWach shows you exactly who's burning what — per user, per repo, per agent, per model — and alerts you before you hit the cap.
No credit card · Free tier covers solo devs and small teams
Your Copilot credit picture in one screen.
Agent sessions that burned the most credits
Repo acme/checkout-service used 72% of monthly credits
Burn rate is 1.6× last month. At this pace, the repo hits its monthly cap by day 22 — and the next 8 days will fall back to the free tier model.
Copilot billing isn't a flat seat fee anymore.
Premium-request credits
Copilot Business and Enterprise now meter premium models (GPT-4.1, Claude 3.7, Gemini) as credits — not unlimited chat.
Per-model multipliers
Different models cost different amounts of credits per request. Same prompt, different bill, depending on the model selector.
Overage billing
Run past your seat allowance and you pay per additional premium request. Easy to blow through in a single afternoon of refactoring.
AI credits are unpredictable by design.
- One engineer using copilot-edit on a big file can burn a week of credits in an hour.
- Agent mode (multi-step refactors) silently fans out into dozens of premium requests per task.
- GitHub's native dashboard updates after the fact — you see the spike in tomorrow's report.
- Switching the default model from GPT-4o to GPT-4.1 quietly doubles the per-request credit cost.
- There's no built-in per-repo or per-team budget — overruns show up at the org level only.
- Free-tier fallback degrades the developer experience right when deadlines are tightest.
Usage broken down four ways, in real time.
By user
Per-developer credit burn. Spot the engineer who switched to GPT-4.1 and forgot to switch back.
By repo
Per-repo monthly totals with trend vs last month. Catch a refactor-heavy repo eating the team's budget.
By agent / session
Group requests into agent sessions — copilot-edit runs, chat threads, agent-mode tasks — and rank by spend.
By model
GPT-4.1 vs Claude 3.7 vs Gemini vs base. See which model is doing the damage and price it in dollars, not credits.
Hear about overruns before the credits run out.
Soft cap warnings
Get pinged at 50%, 75%, and 90% of monthly credits — at the org, team, repo, or user level.
Burn-rate forecasts
AgentWach projects your month-end credit usage from the last 7 days and warns when you're trending over cap.
Hard stops (optional)
Route a webhook to your access-management tool to revoke premium model access for a user or repo when a hard cap is hit.
One Copilot report. Three audiences.
Solo developers
A simple weekly digest: credits used, credits left, projected overage. Free tier, no setup beyond a personal token.
FreeAgencies
Per-client breakdowns with white-label PDF exports. Bill credits back to clients accurately, not by gut feel.
White-labelTeams & enterprises
Org-wide rollups by team, repo, and cost center. SSO, audit log, and per-team budgets with hard stops.
SSO + auditConnect Copilot in minutes.
Paste a GitHub token, pick the org, and AgentWach starts pulling per-user credit data on the next poll. No code change. No agent installs.