Evaluation runs
Read every event of one evaluation run, in order.
GET /telemetry/runs/{runId}Every event of one evaluation run, in the order it happened.
An evaluation run is a batch of scripted conversations run against an agent to
score it. Its events carry a run_id, and this endpoint returns exactly the
events where that id matches and run_type is eval. Production traffic
that happens to carry a matching id is not returned.
Parameters
| Parameter | Type | Description |
|---|---|---|
runId | path | The run id |
cursor | string | nextCursor from the previous page |
limit | integer | Page size. Default 25, maximum 100 |
Response
{ items, nextCursor }, where each item is an event in the shape documented
under events. Every item has
run_type: "eval" and the run_id you asked for.
curl -H "Authorization: Bearer $SUBCORE_API_KEY" \
"https://api.subcore.ai/telemetry/runs/run_2026_09_03_nightly?limit=100"When it 404s
{ "error": "Eval run not found", "code": "NOT_FOUND", "timestamp": 1788566031707 }As with sessions, a 404 covers both "this organization has no such run" and
"this key's permissions.cores does not reach the cores it ran on". An empty
page reached past a cursor is not a 404.
The run's events are spread across the sessions the run created — each scripted
conversation gets its own sid. To follow one of those conversations on its
own, take its sid from an event here and call
GET /telemetry/sessions/{sid}.