TeamCity 2026.2 makes pipelines generally available and adds BYOK for AI Assistant

TeamCity 2026.2 removes the early-access label from pipelines, making them a stable alternative to classic build configurations. The change means teams no longer need to opt in to an experimental feature; pipelines are now enabled by default for new and existing projects. Adoption costs are minimal for users already familiar with YAML-based workflows, but teams that have built extensive build chains around classic configurations may face migration effort if they want to consolidate on pipelines.

The general-availability release adds branch-aware editing, protected-branch warnings, and the ability to run jobs even when upstream steps fail. These improvements address common pain points in CI/CD workflows, particularly for teams working with multiple Git branches or requiring cleanup steps to execute regardless of build status. However, pipelines still lack some build-feature parity with classic configurations, so teams relying on those features may need to maintain hybrid workflows until the roadmap is complete.

Bring-your-own-key support for the AI Assistant shifts control of LLM usage from JetBrains to the end user. Teams can now enforce their own security policies, cost limits, and model preferences without being locked into a single provider. The addition of OAuth/PKCE authorization removes the need to manually generate and manage personal access tokens, reducing operational overhead for AI-driven pipeline management. However, the BYOK model also means teams are responsible for their own API costs and rate limits, which may require additional monitoring and budgeting.

New MCP tools for pipeline management, teamcity_pipeline_get, teamcity_pipeline_post, and teamcity_pipeline_delete, extend AI-driven automation to pipelines. These tools are restricted to Brave mode, which must be manually enabled, reducing the risk of accidental pipeline modifications. The tools complement existing MCP endpoints for build configurations and projects, allowing teams to standardize AI interactions across their entire CI/CD setup. However, the need to manually enable Brave mode may slow adoption for teams that prefer fully automated workflows.