Runs

Reading a run, interpreting citations and mentions, re-analyzing, and exporting.

A run is a single execution of a monitor — one prompt sent to one model at one point in time. Runs are immutable: once written, they’re the historical record. To compare snapshots over time, you query the run history rather than mutating individual runs.

Anatomy of a run

The run detail view has four parts:

The answer — the raw model output, rendered as the user would see it. We don’t trim or reformat.
Citations — every URL the model trusted as a source, with a normalized domain, the anchor text it used (when available), and a match against your tracked brands.
Mentions — every brand or product named in the answer. Your own brand is highlighted; competitors are grouped beneath.
Metadata — model + version, run time, latency, token usage.

Citations vs. mentions

A common point of confusion:

A citation is a URL. The model said “according to nike.com/…”.
A mention is a brand-name string in the prose. The model said “Nike makes good shoes” — no URL needed.

Both matter. Citations tell you which pages are influencing the answer. Mentions tell you which brands are top-of-mind for the model. Share-of- voice rolls up mentions; the content-gap engine works mostly off citations.

Re-analyzing a run

If you change a brand-book entry (e.g., add a product alias the model should have matched), you can re-analyze a run to apply the new rules without re-querying the model. This is fast and free — it’s post-processing, not a new model call.

Exporting

Each run has a markdown export (the raw answer plus citation list) and a JSON export (the full structured payload). Use markdown for sharing in Slack or briefs; use JSON if you’re piping into something else — though the API is generally a cleaner integration point than scraping exports.

API

GET /v1/projects/{projectId}/runs
GET /v1/projects/{projectId}/runs/{runId}

The list endpoint is paginated with a cursor. The detail endpoint returns the full run including citations and mentions inline.