Yesterday, Sam Alba and Andrea Luzzardi announced that the Mendral team is joining Anthropic. They are winding down their hosted AI DevOps engineer, transitioning customers, and heading to work on Claude Platform.
The speed is what gets me. If memory serves, Mendral was first mentioned publicly in January of this year and emerged from Y Combinator’s Winter 2026 batch in March.
Their announcement does not say much beyond the fact that the team is “joining” Anthropic, but given their work and backgrounds, it certainly sounds like an acquihire. Sam was Docker’s first employee and later its VP of Engineering. Andrea wrote Docker’s first lines of code and played a central role in its early growth. This was also their second trip through Y Combinator, following Dagger, the “DevOps operating system,” they co-founded.
Mendral was building an increasingly capable DevOps agent on top of Claude. Now the team is joining Anthropic to help build the platform underneath it.
Talking to Sam about Mendral—and Dagger
In May I sat down with Sam at DevOpsDays Austin to get an update on Mendral and see a live demo. It was our second interview of the year. The first came in March, just after Mendral emerged from YC, where we discussed the company’s origins, where software delivery is headed, and what it’s like to go through Y Combinator twice. [Watch that interview here.] Going back further, we first spoke at KubeCon 2024, when Sam was still at Dagger. [Watch it here.]
When we recorded our interview in May, Sam and Andrea had increased Mendral’s headcount by 50 percent (which meant hiring one person). They had also raised seed funding and signed 30 paying customers. Not bad for a six-month-old company that had recently grown to three employees.
Mendral the product
Mendral was an AI DevOps agent that monitored CI/CD pipelines and acted on what it found.
It was not simply a dashboard that flagged flaky tests. In the demo, Sam showed an intermittent test failure being detected, investigated, and fixed, with Mendral opening a PR in GitHub for approval. He also demonstrated how Mendral could connect error reports with CI logs and recent code changes to trace a problem from symptom to source.
While the demo showed only one agent at work, Mendral shipped three out of the box: security, reliability, and performance. Customers could also define their own for anything specific to their stack.
Claude-powered from day one
Claude had been under the hood since Mendral’s first commit. In a February engineering post, the team explained that they were running Haiku, Sonnet, and Opus all in production, with Opus doing the heavy lifting. Mendral built a pipeline that processed billions of CI log lines each week, providing Opus with the logs and failure history it needed to diagnose why a build had failed and then write the fix.
Building so closely on Claude gave the team a front-row view of how quickly the models were improving. As Sam and Andrea wrote in their announcement: “Every few months, a new model made part of our roadmap unnecessary and a bigger part of it possible.”
That may be the clearest explanation for this move. They were building on Claude. Now they’re joining the team that builds Claude. As they put it, what better place to work on what software engineering is becoming?
Near the end of our May interview, I asked Sam what would be different the next time we spoke. He predicted more custom agents and more integrations. (He did not mention the part about joining Anthropic. 😉
Congratulations to Sam, Andrea, and the Mendral team. I look forward to the next interview.
Pau for now…
Leave a Reply