What changed
AI mention verification — the pass that checks a keyword match is really about your keyword before it reaches you — now runs on a visible credit pool:
- See your balance — the AI-credits meter shows in the dashboard, and agents get the same numbers over MCP via
get_usage. Trial accounts include 1,000 credits a month (up to 500 a week); the paid plan includes 10,000 a month (up to 3,500 a week). One verification costs one credit. - Verification now cleans reads, too — verdicts apply when you or your agent read the Dataset, not only when alerts fire. Records that failed verification are filtered out of keyword-match reads, and results carry their verification status.
Why it matters
Verification quality was invisible: you saw fewer false alerts, but couldn't see the work behind it, and rejected matches still appeared when browsing your Dataset. Now the same "is this really about your keyword?" judgment applies everywhere consistently — and you can see how much of the pool your keywords consume, which is an early signal when a keyword is too broad.
How to use it
Nothing to configure. The meter is on the dashboard's usage panel, get_usage reports it to agents, and Dataset reads of keyword matches are filtered automatically. If a keyword burns credits fast, tighten it — exact phrases match less noise.