README and docs polish

docs: reorder screenshots — runs feed before run detail

Made-with: Cursor

Update docs

chore: update license attribution to AgentMeter

Made-with: Cursor

AgentMeter org change

chore: update all references from foo-software to agentmeter org

Made-with: Cursor

Updating docs

docs: remove early access warning banner

Made-with: Cursor

v1.0.0 — Initial Release

Initial release of AgentMeter Action.

Track AI agent token usage and cost directly in your GitHub workflows. After each agent run, AgentMeter records token counts, model, duration, and status — then posts a live cost summary comment on the triggering PR or issue.

What's included

• Inline mode — drop it after any agent step (claude-code-action, codex-action, or custom) in the same workflow
• Companion mode — attach to any workflow_run-triggered workflow; set workflow_run_id and timestamps, trigger number, status, and tokens all resolve automatically
• Auto token extraction — parses Claude Code JSON output (usage, num_turns), Codex rollout JSONL, and plain-text regex fallback
• Auto turns extraction — extracts turn count from agent stdout without manual wiring
• PR/issue cost comment — upserted on every run, newest-first, collapsible history beyond 5 runs
• Per-model pricing — fetched live from the AgentMeter API with prefix-match fallback for versioned model IDs
• Custom statuses — supports needs_human and any other value alongside the standard GitHub conclusions
• Never fails your workflow — all errors are warnings, not failures

Requirements

• An agentmeter.app account (free, GitHub OAuth)
• AGENTMETER_API_KEY repository secret

Usage

- uses: foo-software/agentmeter-action@v1
  if: always()
  with:
    api_key: ${{ secrets.AGENTMETER_API_KEY }}
    status: ${{ steps.agent.outcome }}
    model: claude-sonnet-4-5
    agent_output: ${{ steps.agent.outputs.output }}

See the README for full usage examples.