Glubean Cloud

Your runs, explained.

Add --upload and every run becomes dashboards, history, load gates, and structured failure evidence — readable by your team and your agent.

Every run, on a dashboard

Pass-rate trends, load pulse, and flaky spotlight — per target.

app.glubean.com — Targets / Default · Overview

Target overview — pass rate trend, load pulse, flaky spotlight

Performance · SLA gates · capacity → p95

Performance — load SLA gates and capacity curves

Runs · test + load history

Runs — test and load runs with status and health

An agent that reads the dashboard

Ask about the run on screen. It answers about that run.

The agent reads structured evidence — not raw logs — then diffs runs, explains regressions, and edits environments, data, specs, and flows. Every write shows a diff before it lands.

Agent · docked beside the dashboard

In-app agent answering questions about runs and editing workspace resources

Diffs two load runs and explains the latency shift
Answers “did error rate get worse?” from real runs
Edits env, data, API specs, and flow steps — with a diff
Never sees your secrets

Evidence, not logs

A failure object says where, why, and what changed.

Cloud keeps the request, response, assertion, and trace as structure — readable enough for people, structured enough for an agent to repair against without quietly weakening the test.

endpointfailed steprequest / responseexpected / actualtrace

failure object

{
  "status": "failed",
  "testId": "auth.get-me",
  "step": "GET /users/me",
  "reason": { "kind": "assertion",
              "expected": 200, "actual": 401 },
  "events": [
    { "type": "http.request",  "url": "/users/me" },
    { "type": "http.response", "status": 401 },
    { "type": "assert.failed",
      "path": "status", "expected": 200, "actual": 401 }
  ]
}

One evidence layer, many readers

The whole team reads the same run — no code required.

Developersdiagnose failures from structured evidence, not CI logs.

Agentsread the dashboard and suggest what to repair next.

QAreview failures without reproducing locally.

Productcheck API behavior without reading tests.

Supportshare run evidence instead of screenshots.

Trust note

Secrets stay local. Evidence goes to Cloud.

Environment values and token secrets never enter the agent context.

22 redaction rules run before any run reaches the cloud.