Free for individual repos

Know what every AI agent run costs.

AgentMeter tracks token usage and cost for every AI agent run — in GitHub Actions CI and local coding sessions from Claude Code, Cursor, and Codex — in one dashboard. Per-run costs traced to issues and PRs. Local session costs traced to the developer and project. No infra. No config files.

View on GitHub Marketplace

By signing in you agree to our Terms and Privacy Policy.

AgentMeter cost summary posted as a GitHub PR comment

AgentMeter runs feed showing cost, engine, status, and duration per run

AgentMeter run detail showing full token breakdown and cost

AgentMeter team settings showing connected repos and member management

Cost surfaces directly in your PRs — no dashboard required.

How it works

From zero to tracking agent costs in under 5 minutes.

Install for CI

Add the AgentMeter Action from GitHub Marketplace to your existing GitHub Actions agent workflow. Works with Claude Code, Codex, and GitHub Agentic Workflows (gh-aw) — no changes to your agent config. Three lines of YAML.

Track local sessions

Use your AgentMeter API key with the CLI to report local Claude Code, Cursor, and Codex sessions. Cost is calculated server-side — no API keys sent. Same dashboard, same metrics, whether you're coding locally or in CI.

Auto-report

CI runs automatically post a cost summary on the PR or issue. Local sessions appear in the feed with project path and device name. No manual reporting — everything is captured.

See

Open the dashboard and see every run and local session with full cost breakdown. Filter by source, engine, or contributor. Spot expensive failures. Set budget alerts before you overspend.

Add in three lines

Drop this step after your existing agent action. No changes to your workflow logic.

.github/workflows/agent.yml

- name: Report agent costs
  uses: agentmeter/agentmeter-action@main
  if: always()
  with:
    api_key: ${{ secrets.AGENTMETER_API_KEY }}
    model: claude-sonnet-4-5
    status: ${{ steps.agent.outcome }}