JetBrains released TeamCity 2026.1, and the headline additions are aimed less at human operators clicking through a UI and more at giving both developers and AI coding agents a direct, scriptable way to interact with CI/CD state: a free, open-source TeamCity CLI with more than 60 commands for investigating failed builds, applying fixes, configuring pipelines, and retriggering builds from the terminal, plus native Model Context Protocol support so external AI tools can query build status, analyze failures, and interact with pipelines through a standard interface rather than a bespoke integration. The CLI ships with what JetBrains calls an agent skill specifically so coding agents like Claude Code or Cursor can invoke it directly, watch build state changes and streaming logs in real time, and reason about a failure without a human first summarizing the build log for it. That matters practically because one of the more tedious parts of AI-assisted development has been the gap between "the agent wrote code" and "the agent knows whether that code actually built and passed CI," which historically has needed a human to paste log output back into a chat window. Native MCP support closes that loop by default rather than requiring a custom integration per tool. The release also brings YAML- and now Kotlin-DSL-defined Pipelines further out of early access with upstream dependency support, letting large monorepo or microservices setups decompose one giant pipeline into smaller, independently maintainable ones that can still be chained together, alongside bundled SAML SSO support and short-lived, build-scoped GitHub tokens that avoid handing long-lived credentials to every build step.