Skip to Content
webagent docs
Concepts

Evidence

What proof a run leaves behind, how to reach it, and how long the links last.

Open as .md

Evidence is what you show someone after a run: a video, a log, a report. This page lists what exists today and how to get it. Some of it is only available through the REST API.

What exists

EvidenceHow to get itNotes
Live viewDashboard, Sessions, open the sessionOnly while the session is open. See Live view
Share linkPOST /v1/sessions/:id/shareLive or replay, no API key for the viewer. 60 seconds to 7 days
Replay videoGET /v1/sessions/:id/recordingOnly if the session was created with record: true. Available after the session closes
Event logGET /v1/sessions/:id/logs (JSON pages), or GET /v1/sessions/:id/logs.ndjsonConsole messages, page errors and network events of the session. Available after the session closes, until retention removes it
Test reportsmoke_test or get_report (MCP), GET /v1/skill-runs/:id/report.json or GET /v1/tasks/:id/report.json (REST)One JSON report: verdict, steps, findings, console errors, failed network calls and links to the other evidence. See Reading results
Skill report (HTML)reportUrl in the skill result, evidence.reportSignedUrl in the test report, or GET /v1/skill-runs/:id/reportAn HTML report with the steps and findings. Failed steps may carry a screenshot, for example in smoke-uat
Task resulttask in the run_task and get_task resultAnswer, stop reason, steps and cost. No screenshots

A task result has no screenshot, page URL, console log or network log of its own. get_report with the task id adds the console errors, the failed network calls and the page addresses, read from the session log. A smoke_test is always recorded. For a goal run, give run_task record: true to get a recording. To share one browser over several tasks, open the session with create_session and record: true, run the tasks with sessionId, then call stop_session. The recording and the log are stored when the session closes.

The session log and the report

Console errors and failed network calls in a test report come from the session log. The log is saved when the session ends. A report read before that has empty lists and a note that the log is not available. Read it again a few seconds later. A task that left its session open has no log until you stop the session.

Fetch evidence

API address

The shell examples on this page use the variable API_URL. Set it once: export API_URL=https://api.us.webagent.lookover.ai. If your organization was given another API address, use that one instead. Set WEBAGENT_API_KEY to your API key.

The recording and log routes answer with a redirect to a short-lived file URL. Follow the redirect:

Download a recording
curl -L -H "Authorization: Bearer $WEBAGENT_API_KEY" \
  $API_URL/v1/sessions/$SESSION_ID/recording -o recording.webm

The file is a WebM video.

A session without a recording, or one still open, answers not_found.

LinkLifetime
Recording, thumbnail and log redirects5 minutes
Skill report from GET /v1/skill-runs/:id/report, and reportUrl5 minutes
Links inside a test report (evidence.reportSignedUrl, steps[].screenshotUrl)5 minutes. The report says when in evidence.urlsExpireAt. Call get_report again for new ones
Share linkWhat you set with ttlSeconds. 60 seconds to 7 days, default 1 hour

These are the link lifetimes today and they may change. The 5 minute links are for opening or downloading at once. Ask for a new one when you need it again. If you want to keep a file, download it.

Stored evidence is not kept forever. The event log is available until retention removes it, and this page does not state when. Download what you must keep.

On this page