See what your agents did, and what it cost.

• Runs as they arrive

  The live screen lists each run as its events land and refreshes every few seconds: the agent, its status, model calls, tokens in and out, and cost. Open a run to see every event inside it.

• Cost you can check

  A model call costs what you send with it, or our rate for a model we know. A model we cannot price counts as $0 and the run is flagged as unpriced, rather than guessed at.

• Keys and a monthly cap

  Each tier sets how many keys you can hold and how many events a month it accepts. A key is shown once, stored as a hash and revocable. Past the cap, ingest refuses with a plain message rather than charging for overage.

• Watches, never controls

  There is no pause, no kill and no approve anywhere in the console. It records what your agents report, and nothing it shows can change a run.

await fetch('https://review-my-site.com/api/agents/ingest', {
  method: 'POST',
  headers: { authorization: 'Bearer agk_...', 'content-type': 'application/json' },
  body: JSON.stringify({
    run: { id: runId, agent: 'support-bot', environment: 'prod' },
    events: [
      { seq: 1, kind: 'start' },
      { seq: 2, kind: 'model_call', name: 'plan', model: 'claude-sonnet-5',
        input_tokens: usage.input_tokens, output_tokens: usage.output_tokens },
      { seq: 3, kind: 'end', detail: { status: 'succeeded' } },
    ],
  }),
});

Without a subscription you get 1 key and 10,000 events a month. Watch takes 100,000 events a month. Audit raises that to 500,000, and Govern to 2,000,000.

Free is the absence of a subscription, not a trial. It does not expire, nothing auto-converts, and doing nothing charges nothing. Every figure above is read from the same catalogue the checkout charges against, so a price shown here cannot differ from a price taken.

• Step by step replay

  One playhead over a run, with a lane per agent, the diff each step wrote, and spend and tokens adding up as it plays.

• Flag a step back to the agent

  A brief drafted from the step you flag, collected by the agent over MCP and marked as raised by a human.

• Policies and budget alerts

  Rules checked after a run finishes, and an alert when an environment passes its budget. Advisory: neither would stop a run.

• AI reviewers and a copilot

  Reviewers over a finished run that publish their hit rates so you can switch one off, and a copilot that cites the step behind every claim.

• Runtime attribution

  Per-step tokens and tool arguments from Claude Code, commands and exit codes from Codex, and run totals from anything else.

export function CouponField({ onApply }: Props) {
  const [code, setCode] = useState("");
  const [error, setError] = useState<string | null>(null);

  async function submit() {
    await onApply(code);
    if (!code.trim()) return setError("Enter a code first.");
    const res = await onApply(code);
    if (!res.ok) return setError(res.message);
    setError(null);
  }