Momentic Launches Mo, an AI Agent for Scriptless Software Testing

Momentic has launched Mo, an artificial intelligence agent designed for scriptless software testing. The company announced the release on September 28, positioning Mo as a direct answer to a bottleneck that has grown sharper as AI coding agents flood teams with machine-written code.

The pitch is simple: give Mo an objective and access to a web, iOS, or Android app, and the agent does the testing itself. What comes back is a report containing discovered test cases, the bugs it found, coverage gaps, screen recordings, and step-by-step reproduction instructions for each defect. No test scripts are written, stored, or maintained anywhere in the process.

How Mo works: a swarm instead of a script

Mo looks familiar to anyone who has used an AI coding assistant. It ships with an agentic harness, a web interface with a chat box where a developer types what needs testing and pastes a URL, and a command-line interface for automating the process.

Behind that interface, Mo spins up a swarm of specialized agents. Exploration agents map out the test cases, bug-bash agents hammer through them, and a reproducer agent confirms suspected defects before anything gets filed. The agent can also ingest extra context when a feature needs it, including a product requirements document, a Jira issue, a Linear ticket, or a Confluence page.

The swarm then attempts thousands of permutations and edge cases while recording every interaction. When the run finishes, the developer gets a blow-by-blow account of what was tested, what failed, how to reproduce each bug, and video evidence of every one. According to Momentic's product documentation, Mo can run one hundred or more agents concurrently, and a typical run can move from a URL to a filed report in about twenty minutes.

"There are no tests in the future"

Momentic's argument starts with a familiar complaint. Test scripts that click buttons, trigger drop-down menus, and fill in fields work until the application changes. Then they break, and the suite itself becomes a liability that developers must check, modify, and fix every time the app evolves.

"As long as you have scripts somewhere in your codebase, someone has to maintain it," co-founder and chief executive Wei-Wei Wu told SiliconANGLE in an interview. Wu's longer-term bet is more radical: "What is the next frontier of agentic quality? Our bet is that there's no tests in the future. It's just agents verifying it themselves, based off of your instructions."

The timing matters because AI-driven build cycles have changed how software gets written. A developer may now oversee a coding agent that produces large volumes of code, review it for obvious errors, and then hand it off for testing. But the tests themselves are still code that must be written, maintained, and verified. Scriptless software testing is designed to remove that intermediate layer entirely, letting quality-assurance agents work from requirements and intent directly.

"The scripts are not the end goal," Wu added. "No one goes to work and is like, 'Hey, I want to write ten thousand scripts.' What they want is no bugs."

Agents testing agents' work

The launch builds on a platform update Momentic rolled out in June, which introduced an agentic quality platform running on a shared product knowledge base. That system includes an Explore Agent that automatically expands test coverage and a Failure Classification Agent that categorizes failures and opens pull requests to fix non-bugs.

Early beta results shared by the company suggest the approach scales. The agents analyzed more than seventy thousand test failures, opened four hundred pull requests, wrote six hundred tests with a seventy-three percent merge rate, executed over two billion steps, and verified more than eighty thousand pull requests. Those figures come from Momentic's own beta reporting, as carried by the June announcement.

Momentic said customers have been putting Mo through its paces over the past few weeks. The company names technology and service companies including Notion, Superpower, Iris, Committee for Children, and Boundless among those testing the new agent. Founded in late 2023 and backed by Standard Capital, Y Combinator, and Dropbox Ventures, the San Francisco company has been building toward this launch for the better part of a year.

There are reasons to stay measured. The concurrency and twenty-minute figures are vendor claims rather than independently verified benchmarks. And the deeper question Mo raises is about the role of the quality-assurance engineer: if agents write the code and agents test the code, the human's job shifts from authoring scripts to reviewing agent-generated evidence and deciding what "correct" means.

Still, the direction of travel is clear. Coding agents keep getting cheaper to run, which means the volume of machine-written code will keep climbing. Somebody — or something — has to check it before it ships. Momentic is betting that the something will be a swarm of agents with a chat box on top. As agent infrastructure goes open source, the QA layer may be next to fall.

Sources: SiliconANGLE (primary reporting and Wu interview); Let's Data Science (product documentation details); dcmetrowire / Reportable (June agentic quality platform announcement and beta metrics).