The news
AI software testing got a new entrant on September 28, 2026, when startup Momentic launched Mo, an agent that tests running applications without a written test suite. Coding-agent company Blitzy, meanwhile, described mapping enterprise codebases in graphs so its agents understand what they change.
Developers give Mo an app URL and test credentials through a chat window or command-line interface, and it sends a swarm of agents to click through the app and try thousands of permutations and edge cases, SiliconANGLE reported. Mo returns reproducible bug reports with video. Co-founder and CEO Wei-Wei Wu told SiliconANGLE: “Our bet is that there's no tests in the future.”
Momentic's product page offers Mo through free self-serve sign-up for web, iOS and Android apps, running 100 or more agents per session. Pricing is usage-based credits, with $250 in starting credit for self-serve sign-ups. Momentic is pitching Mo at teams whose coding agents produce changes faster than engineers can review them and that have no dedicated QA owner. Momentic also claims Mo has caught 117,010 bugs before deployment. SiliconANGLE reported that Notion, Superpower, Iris, Committee for Children and Boundless had been testing Mo in the weeks before launch.
Blitzy tackles the other end of the problem. In an interview at GraphSummit 2026, reported by SiliconANGLE on September 28, Blitzy director of engineering Neeraj Deshmukh said the company reverse-engineers customer environments and builds codebase graphs on Neo4j's graph database. The graphs connect to GitHub and GitLab and update when code changes.
Deshmukh said an agent's effective context tops out at about 200,000 to 300,000 tokens, roughly 20,000 to 30,000 lines of code, and that on a 100-million-line codebase agents start summarizing and dropping information. According to SiliconANGLE, Blitzy cited an 84.95% score in June 2026 on SWE-Bench Pro, a benchmark of real-world software engineering tasks.
Blitzy raised $200 million at a $1.4 billion valuation in a round led by Northzone, Crunchbase News reported on May 5, 2026. The company said then that dozens of Global 2000 enterprises across 10 industries use its platform, naming State Street and QAD.
The numbers
- Bugs caught before deployment (Momentic claim)
- 117,010
- Mo starting credit for self-serve sign-ups
- $250
- Context per Blitzy agent (company estimate)
- 200,000-300,000 tokens, about 20,000-30,000 lines of code
- Blitzy SWE-Bench Pro score cited by the company, June 2026 (per SiliconANGLE)
- 84.95%
- Blitzy funding round, May 2026
- $200 million at $1.4 billion valuation
Why CEOs should care
For CTOs and engineering leaders, our reading of both companies' pitches is that generating code is no longer the constraint; checking it and fitting it into existing systems is. If coding agents have raised your team's output, look at whether testing and code review capacity grew with it. Ask testing vendors how often their bug reports reproduce, how many findings turn out to be false alarms, and how credit consumption scales with app size.
For CISOs, an agent that logs into your app with test credentials, or a platform that reverse-engineers your codebase into a graph, needs the same review as any vendor with deep access. Use dedicated test accounts and non-production environments, and ask where the codebase graph is stored, who can query it and how it is deleted when a contract ends.
For CFOs, both products shift spending rather than remove it. Momentic prices Mo on usage credits, adding another metered AI line to the engineering budget. Before approving coding-agent spend, ask engineering leaders what share goes to verification and codebase context, and what measure, such as escaped defects or production incidents, will show whether it paid off.
The bigger picture
Investors have been paying up for AI coding. Crunchbase News noted in May that comparable companies included Anysphere, valued at $29 billion, Replit at $9 billion and Lovable at $6.6 billion. Deshmukh framed the next problem as whether AI understands the system it is modifying, not whether it can write code. That puts pressure on the tools around code generation, including testing, review and codebase mapping, which is where Momentic and Blitzy are placing their bets.
What’s next
Watch for independent evidence on AI testing agents, such as customer data on escaped defects and false positives, and for published per-credit rates beyond Momentic's starter credit. Also watch whether coding-agent vendors start bundling testing and codebase context into their own platforms, which would make stand-alone tools harder to justify.
What “Fact-checked” means
Fact-checking means testing a story’s facts against the evidence before it is published. This story went through at least two separate checks before this version was published.
- What we checked
- Its names, figures, dates, job titles, quotes and who said what were checked against the story’s sources, including its main source where it could be opened. The headline was checked for accuracy and overstatement.
- How
- A first check reviewed the whole story. If it passed, a second, skeptical check went back to the sources to look for mistakes in the most important facts. If a check flagged the story, it was edited to fix the problems found, and a separate re-check then reviewed the whole story again.
- Who
- The checks are made by our newsroom, as steps kept separate from the writing, under rules set by our editor, Hussein Mukhtar. A story the checks still flag is held for the editor, who decides whether it is fixed, published or dropped.
- If something is wrong
- “Fact-checked” does not mean error-free. If a material error is found after publication, we correct the story and add a note saying what changed. Report an error
