Status

Controlled free beta is live. The public MCP endpoint and its production database are running, while payments and public indexing remain off. The first production scheduled-evidence run has not yet been observed, and no store has connected, so those two real-world results are not claimed.

• Measured accuracy: none. No connected store has confirmed a cause yet, so diagnostic accuracy is genuinely unmeasured — see EVAL-REPORT.md, where it is recorded as UNMEASURED rather than estimated.
• Real store pilot: not run. The collectors have never executed against a real WooCommerce install.
• Scheduled collection in production: configured but not yet observed. The cron is written and tested; this page will report its recorded state after its first real run.
• Payments: off. Every tool answers free.
• Indexing: off. Every page is noindex until a human approves it on a real domain.

Live operational state

Nothing fresh enough to diagnose from. A diagnosis needs data inside its window. Without it the tools abstain rather than answer from what is left.

• Public evidence refresh: fresh.
• Store snapshot ingest: expired. A grade rather than a count: with one store connected, an upload count would be that store’s activity.
• Reason codes: SNAPSHOT_EXPIRED.
• Operations policy: version 1.0.

The two pipelines are reported separately on purpose. Public evidence collection succeeding says nothing about whether any store’s snapshots are arriving, and one standing in for the other is how a broken half stays invisible.

What each state means

• NOT_CONNECTED — no store has sent a valid signed snapshot. Not a fault and not health; nobody has taken the step yet.
• COLLECTING_WITHIN_POLICY — the observations a diagnosis needs are inside their policy windows and both pipelines are healthy.
• PARTIAL_EVIDENCE — something a diagnosis would lean on is truncated or limited. Answers still come, with the limit named on them.
• STALE_OR_PIPELINE_DEGRADED — the last valid data is inside its warning window, or a pipeline has failed enough times in a row to count.
• NO_USABLE_EVIDENCE — connected once, nothing fresh enough to diagnose from now.
• STATUS_UNAVAILABLE — the policy or the database could not be read. It is never used for “nothing here”; that is what the first two states are for.

Coverage for any single answer is reported on that answer, in its signal-family count and independent-source count. There is no separate number here that would not be redundant with what every answer already states.