Momentic raises the bar for software testing with agentic quality platform

Artificial intelligence-powered software testing and quality assurance platform Momentic Inc. today announced a major update to its service focused on verification in the AI coding era, enabling teams to ship more code.

Quality assurance serves as a stopgap at the end of code production, preventing bugs from reaching production. Although much of the coding experience involves writing quality code from the outset, the advent of AI-driven development lifecycles has introduced not just swifter build speeds, but also a whole new species of problems.

“AI adoption is scaling, and especially in engineering, everyone’s using tools like Cursor, Claude, Codex, any of the plethora of these coding agents to help them write more code,” co-founder and Chief Executive Wei-Wei Wu told SiliconANGLE in an interview.

It goes without saying that software quality remains paramount, especially in the AI era, where agentic tools now do 40% to 50% of the work and are now exponentially scaling up code output. As a result, they’re also exponentially scaling up bugs.

Recent industry research suggests that AI coding tools are accelerating development, but not eliminating the need for oversight. GitLab’s 2026 Global DevSecOps report described this as an “AI Paradox,” finding that faster coding is creating new bottlenecks around quality, security and compliance, with only 37% of respondents saying they would trust AI to handle daily work tasks without human review. Momentic’s pitch is aimed directly at that emerging gap: If AI agents are going to write more code, teams need faster ways to verify whether that code still works.

“Quality is extremely important in this world,” Wu warned. “Where ‘AI slop’ is very prevalent, you’re just vibe coding to production.”

Making quality assurance part of the culture

The new Momentic offering covers a number of gaps. It starts with what Wu called the Explore Agent, which covers knowledge gaps. It moves on to a Failure Classification Agent that automatically triages and categorizes failures to determine if something needs handling or not. Finally it settles on making tests intent-based and readable.

Explore Agent acts as an approach to agentic memory that can watch everything. It takes scattered information from every part of the business, not just the codebase, including Jira, Linear, Figma, Slack, Zendesk and what developers say about the code.

“The more you use Momentic, the smarter it gets,” Wu said. “It’s able to understand, hey, these are areas of your code base or your product that are not covered by anything; you’re kind of shipping blind.”

However, test coverage is only half the battle. When tests run, they’re just an indicator that something works or it doesn’t. Developers don’t always have a clear idea if that’s an example of quality. That’s where the Failure Classification Agent steps in. Sometimes tests are “flaky,” which means they’re not written quite right, or the bug they’re checking for isn’t actually a bug; it’s just something that looks wrong but isn’t.

The industry buzz phrase people are used to hearing is “signal-to-noise ratio.” If something failed because the original test wasn’t updated under the hood after the user interface was changed, then it’s the test that broke, not the code. If the test failed because of a warning, but it’s not a security vulnerability and it will not explode in production, this can wait in line behind a critical fix.

“If it’s an intended UI change and the test failed, Momentic triage can automatically update it for you,” Wu noted.

Having a guide as to the level of severity of a given test failure means that developers can look at the quality of their code before preparing to push merges to production during review. This also reduces the number of code reviews that are kicked back to the original developers based on misunderstood tests.

And that means the tests themselves need to be human-readable, not just machine code.

Make the tests work for developers, not the other way around

One interesting outcome of the agentic and vibe coding era is that something that almost every development team does almost as naturally as breathing is becoming a must-have: creating specification documents before going forward. The specification — that is, the plan document — has become the point of truth for AI agents to build guidance before going to work.

Naturally, that specification should inform the test suite. If the end product should behave as specified, then the tests should represent English-readable formats that speak to the how and why of the behavior expected from the underlying code.

“The test format is a plain-English description of what you have built and what are the things that you care about, whether it’s happy paths, edge cases or things like that,” said Wu.

That allows developers to easily understand what’s being tested and why — and, most critically, update it with clear instructions. It also allows them to collaborate with AI agents in a similar way they would with another employee. Even better, the tests themselves are readable by other employees who themselves will need to understand them to maintain them.

Image: SiliconANGLE/Microsoft Designer