Hover Cloud

Hover Cloud is the hosted layer over the specs you already own. It ingests your CI results and gives you trends, flakiness, regression alerts, and self-heal orchestration — a data and control plane, not a test runner.

Your specs keep running where they already run: your own GitHub Actions, free minutes, plain Playwright, zero AI in the green path. Cloud never executes a browser. It reads what your CI reports and acts only on what already ran.

The lines Cloud won't cross

These aren't limitations to fix later — they're the product:

  • Cloud runs no browsers. Execution is your CI. Cloud is a data product, not a compute product. This is both the cheapest thing to run and the strongest anti-lock-in position — we don't even run your tests.
  • AI only on what already ran. The green CI path is 100% AI-free. Cloud's AI touches a failed or flaky spec — to diagnose, to route a heal, to summarize a PR — never a green build, never your local authoring.
  • No lock-in. The specs are plain Playwright in your repo. Delete your Cloud account and they keep running in your CI, untouched.
  • Heal fixes are local and human-reviewed. Cloud detects the drift and routes a heal request back to your editor; your own agent re-locates it in your real browser, and you review the diff. Cloud never silently rewrites your specs.

How the pieces fit

LayerWhere it runsWhat it does
AuthoringLocal, freeThe @hover-dev/mcp server drives your own coding agent to crystallize specs into your repo.
RunningYour GitHub ActionsThe crystallized specs run as plain Playwright on every PR and on a schedule.
Cloudcloud.gethover.devIngests run results + traces; dashboards, flakiness, regression alerts, heal orchestration, team environments.

Get set up

  1. Connect a repo — sign in with GitHub, pick a repo, and Cloud generates the CI workflow.
  2. Environments & accounts — point the same specs at staging or production, and wire in test logins without leaking secrets.
  3. Runs, alerts & self-heal — trigger runs, read results, get regression alerts, route heals.
  4. Connect your editor — pull the heal queue and cloud dashboard into VS Code and your MCP agent.
  5. Team & lifecycle — invite teammates, set up scheduled monitoring, rotate tokens.