Skip to main content
Mo is an AI QA engineer that bug bashes your app. You give it a target and a brief with the expected behavior, test accounts, and actions to avoid. It explores the app with parallel agents and returns a report of what it checked, what it couldn’t verify, and each bug with reproduction steps.
To try it now, run your first Mo session.

Targets

A session tests one target:
  • A web app at a URL, in a hosted browser: production, a preview deployment, or a local build through a tunnel.
  • A pull request. Mo reads the diff and tests the PR’s preview deployment, passed in or discovered from the PR. See PR sessions.
  • An iOS app on a remote simulator, or an Android app on a remote emulator, installed from a build you uploaded to a channel and tag. See iOS app setup and Android emulators. Mobile sessions require Mo mobile access in your workspace.
Mo doesn’t build or start your app. To cover web, iOS, and Android, start one session for each.

What a session produces

A session keeps one report. It holds:
  • Agent outcomes: verified, issues found, or blocked. Review the summaries and gaps to understand what Mo checked.
  • Findings with expected behavior, actual behavior, and reproduction steps. Functional bugs require independent reproduction. Static findings, such as a typo, use a screenshot.
See session reports for the fields, and triage for the decisions you make on a bug.

Context the session reads

Beyond the brief, a session reads:
  • Workspace repositories you select on the Mo workspace page provide code and diffs through read-only Git access. Mo refreshes default branches at session startup when possible; a cached checkout may be stale if the refresh fails. Name the branch or revision in a brief that depends on a specific change.
  • A named Momentic environment provides variables to the session, and its BASE_URL variable sets the starting URL.
  • An uploaded spec or test plan guides Mo’s checks.
  • Integrations such as Linear, Jira, Confluence, and Notion, plus custom MCP servers your workspace adds, give it tickets and specs.
See run a session for the settings and diff-based regression for the checks across a diff window.

The agents in a session

A session has a session agent and its sub-agents. Each role has one job: For a suspected functional bug, a reproducer agent walks the flow from the beginning in a separate browser or device session. It uses the authentication and setup instructions with equivalent test data. If it cannot reproduce the behavior, Mo does not add it as a confirmed functional bug.

Parallel agents

Every sub-agent gets its own browser, simulator, or emulator, so agents check flows at the same time. Choose effort to match the scope of your brief. Separate browsers and devices can still share accounts and backend data. Provide disposable test data or isolation rules in the brief.

Limits

A web session runs in a hosted browser. Flows that require operating-system dialogs or another desktop app, such as a print dialog or a mailto: link, can block verification. Review the coverage gaps in the report before treating a session as complete.

Keep a finding as a test

When a bug is worth checking on every change, have the Coder write a regression test for it. The test is a YAML file in your repository that runs locally and in CI. See how regression tests work.