Five steps, no test code. Point me at a URL and I take it from there.
Point me at a URL and I explore your app page by page: every route, form and flow, found, named and connected. There is no setup and no test code to write first; the map is where every flow I test begins.
From the map I propose the flows worth testing, written as plain-English steps anyone on your team can read. You approve them, edit them or add your own, and nobody writes test code.
Every test runs in a real browser and you can watch each step in the dashboard as it happens. Start runs by hand, on a schedule, on every pull request, or in your own signed-in tab with the Chrome extension.
When a step fails I stop there, lock on to what broke and write it up: steps to reproduce, expected against actual, screenshots, console and network logs. Everything a developer needs to fix it is in one report.
On GitHub I add a check to every pull request and leave one comment with the failure summary. Claude, ChatGPT and Codex reach me over MCP, and your CI starts runs with the CLI, the API or a webhook.
codex mcp login outqube.OUTQUBE_API_KEY.Add Outqube to the branch's required checks and a failed test blocks the merge. I post one comment per PR and update it on every push.
Point your client at mcp.outqube.com and sign in with Outqube. CI uses an API key.
Twelve MCP tools. Ten read runs, steps, screenshots, logs and failure reports; two start or cancel a run.
I read the commit, branch and PR from GitHub Actions, GitLab CI and Azure Pipelines. A webhook starts a suite after a deploy.
Give me a URL. I'll map your app, propose the flows and start testing.