Mo is an AI QA engineer. Point it at your app on web, iOS or Android and get bugs back, each with a recording and reproduction steps.
Start Mo from your coding agent (Claude Code, Cursor, Codex, Copilot and others). In your project root:
npm install -g qa
qa login
qa skillsqa skills asks which agents to set up and installs the mo-qa skill for
them. Then ask your agent for a bug bash:
/mo-qa bug bash the cart on https://staging.example.com/cart.
Do not place an order.
The agent writes the brief, starts the session, follows it and exports the report. It can also fix the bugs Mo finds and recheck them.
You can also start a session from the app or by
mentioning @momentic-ai on a pull request
(GitHub setup). Mo needs a
Momentic account: sign up. See the
quickstart.
Momentic also runs plain-English end-to-end tests. They live as YAML files in your repo and run locally or in CI.
QA Bench runs Mo, Codex, Claude Code and Devin against 20 product areas of Gitea, Metabase, Twenty and Saleor, then grades their bug reports against 1,396 human-reviewed bug labels. With 60 minutes per area, Mo found about 269 labeled bugs at 84% precision. The next best agent found about 86.
- qa-bench: the QA Bench harness, fixtures and results.
- qa: the CLI for Mo.
- skills: agent skills and MCP servers for coding agents.
- cli: the Momentic CLI. Write, run and repair web tests from your terminal.
- mobile-cli: the Momentic CLI for iOS and Android tests.
- examples: example projects and CI setups.
momentic.ai | Docs | X | Y Combinator W24