Analytics for agent-built apps
Ask your product anything.
One tag instruments your app. The metrics come certified. You ask in plain English, and your agents ask over MCP.
askhow many people actually used the app this month?
example
Unique visitors · last 28 days
24,312▲ 12.4% vs prior 28d
certified metric · receipt one click awayanswered in 34 ms
example
Unique visitors24,312▲ 12.4%
Sessions30,631▲ 10.5%
Pageviews83,400▲ 11.9%
Avg session2m 34s▼ 4.3%
Bounce rate39.1%▼ 1.1 pt
Sessions per day current prior 28d deploys
example
▸ask("did yesterday's deploy hurt checkout?")
INP p75 on /checkout rose 212 ms → 348 ms within 40 min of deploy 4f2c1a9. 1,214 interactions measured.
▸ emit_event("agent.action.deploy", { commit: "9b1f04d" })
recorded · on the timeline next to the metrics it affects
Agents read the same certified truth, and their deploys land on the same timeline as the metrics they move.
Private by design
- First-party only. No third-party requests, ever.
- No fingerprinting. A random id in your own cookie.
- Ids are scrubbed from paths before storage.
- PII is flagged in the contract, and one switch silences the tag.
What we never do with your datav1.0, effective 2026-07-28
- Your events are never used to train a model. Not ours, not anyone else's.
- There is no opt-in by default, and no exception by region, plan, or account age.
- Your workspace is never pooled with another workspace to build anything.
- Changing any line above requires a new version and a new date, in public git history.
01
One tag. Autocaptured events, certified metrics.
02
Ask in English from the dashboard. Agents ask over MCP.
03
Trust is structural: every answer can show its work.
Live in production on seven sites right now, including this page.