AI agents can now configure TeamCity build pipelines end-to-end, not just generate configuration suggestions. Two experiments demonstrate this: in the first, ChatGPT designed a multi-build-chain setup using TeamCity docs via Context7 MCP, then Codex executed it via the TeamCity REST API and CLI, iterating on errors until the configuration worked. In the second, Codex set up CI for a Go project from scratch, diagnosed a missing Go runtime in the build agent, worked around it, and fixed test reporting after feedback. The key insight is the agent's feedback loop: apply configuration, observe results, adjust, and retry — converging on a working pipeline without manual developer intervention between cycles.
Sort: