At BuildBuddy, we've been thinking a lot about what it means to build developer infrastructure not just for humans, but for agents. This post, which was adapted from my BazelCon talk, walks through the tools we've built to make building with agents easier and faster.
Configuring AI coding agents to use Remote Bazel can unlock significant performance improvements for builds and tests. With more compute, a warm analysis cache, and fast network access to co-located remote caches, our remote runners are optimized to ensure agents' time is well spent on logic and reasoning, instead of waiting for builds to complete.
For more details on what Remote Bazel is and some sample use cases, see our blog post on Remote Bazel. For more details on the technical setup of our remote runners, see this blog post.
AI coding agents like Claude, Cursor, and Codex are transforming how developers write software. For many developers coding with agents, the bottleneck has shifted from writing code to validating code written by agents - with builds and tests.
Remote Bazel ensures these builds and tests run quickly by running them on powerful remote runners. These runners are colocated with cache and RBE servers, reducing network latency. They can be cloned and reused for future builds, ensuring a warm analysis cache. All the local machine needs to do is stream back the logs.
It's like giving your AI a build farm.