Proof & evidence engine

The Opsis MCP server is Opsis's proof engine. Findings from replay analysis stay suspected until the MCP server successfully reproduces them in a real browser. Only then do they become verified.

What happens when you click “Run verification”

1. Opsis takes the verification plan from the finding (created by AI from session evidence, or synthesized from signals).
2. Headless Chromium runs against your app's URL, replaying the user's steps: open the route, fill the form (with synthetic test data), click the problematic element.
3. During the run, Opsis records video, a trace, console messages, page errors, failed responses (≥400), and request failures.
4. The bug counts as proven when the expected console error or failed request pattern shows up again.

Verification runs interact with your app at its configured app URL. Point it at a staging environment if you prefer not to run synthetic interactions on production.

Artifacts

storage/artifacts/<project-id>/<run-id>/
  video.webm        # screen recording
  trace.zip         # full trace archive
  screenshot.png    # final page state
  console.json      # console + page errors (redacted)
  network.json      # failed requests (redacted)
  result.json       # structured result

Artifacts are private: they are served only through an authenticated gateway that checks your organization membership. Artifacts expire automatically after your plan's retention window (see Scans & retention).

Run statuses

• passed: the bug was reproduced; evidence is stored; the finding becomes verified.
• failed: the flow ran normally and nothing reproduced. If it wasn't a fix re-run, the finding becomes not_reproducible; a re-run of an already- verified or fixed finding becomes fixed; a re-run after a proposed fix becomes fixed_candidate until you confirm production.
• inconclusive: errors were observed during the run, but none could be attributed to the suspected bug. The finding's status is left unchanged and the captured video and trace are kept for human review.
• error: the runner itself failed; no conclusion about the bug.

Retention & private artifacts

• Artifact retention is tiered by plan: 3 days on Free, 14 days on Pro, 30 days on Team, and 365 days on Enterprise. An automatic sweep removes expired artifacts.
• Fix-evidence grace: the artifacts of the latest run of a finding that is currently verified or fixed survive past retention. The live proof of an open bug (or of a fix that worked) is never swept while the finding still depends on it.

More on scan cadence, retention windows, and digests: Scans & retention.