TeamCity 2026.2: Pipelines General Availability, BYOK for AI Assistant, and More

JetBrains released TeamCity 2026.2, making pipelines generally available for On-Premises and Cloud, and adding BYOK for its AI Assistant, allowing users to connect any major cloud AI provider or local LLM. The update also improves branch handling, adds run-through-failure job dependencies, pipeline promotion, and debug jobs, alongside bug-fix updates for versions 2025.11.8 and 2026.1.4.

TeamCity Powerful CI/CD for DevOps-centric teams News /teamcity/category/news/ Releases /teamcity/category/releases/ TeamCity /teamcity/category/teamcity-2/ TeamCity 2026.2: Pipelines General Availability, BYOK for AI Assistant, and More This major update for TeamCity On-Premises introduces a number of pipeline improvements that are now moving out of EAP and joining build configurations as official, stable options for designing CI/CD workflows. We are also shipping a number of AI-related features, including an MCP toolset extension and the ability to choose any major cloud AI provider or your local LLM to handle all interactions with AI Assistant in TeamCity. As always, you can see the complete list in our documentation https://www.jetbrains.com/help/teamcity/2026.2/what-s-new-in-teamcity.html . We also roll out two bug-fix updates for our previous major TeamCity On-Premises versions: 2025.11.8 https://www.jetbrains.com/help/teamcity/2025.11/teamcity-2025-11-8-release-notes.html and 2026.1.4 https://www.jetbrains.com/help/teamcity/2026.1/teamcity-2026-1-4-release-notes.html . TeamCity pipelines General availability TeamCity 2025.07 introduced pipelines as an Early Access Program EAP feature – instantly available on TeamCity Cloud and on request for TeamCity On-Premises. Since then, we’ve been steadily closing the gap with classic build configurations, adding build feature support, .NET steps, the ability to link pipelines into build chains, and more. With this release, pipelines move out of EAP and are now generally available on both TeamCity Cloud and On-Premises, for projects of any size and complexity. However, our development of pipelines does not stop here – see our roadmap https://www.jetbrains.com/help/teamcity/pipelines-roadmap.html for what’s next. Improved branch handling Pipelines are now much better equipped to handle branches when your YAML file is stored in a Git repository. You can now select repository branches while in Edit mode, and design unique workflows for each branch. In addition, pipelines are now better at handling protected branches – the most common technique for safeguarding your main branches from unverified commits. Starting with version 2026.2, pipelines show clear warnings that edits made in the UI cannot be pushed to protected branches directly, and suggest saving them to other branches instead. Learn more: Feature branches https://www.jetbrains.com/help/teamcity/2026.2/pipeline-settings.html Feature+Branches Run through failures When setting up job dependencies, you can now enable Run job even if upstream fails to ensure a pipeline runs until the very end, even when there are errors. The pipeline run will still be marked as failed overall, but this feature lets you ensure that specific jobs for example, cleanup or notification steps always run. Learn more: Job dependencies https://www.jetbrains.com/help/teamcity/2026.2/job-settings.html Dependencies Promote pipeline runs Promote – the button that triggers the downstream part of a build chain from an older, already finished build – now works for pipelines, not just build configurations. For example, you can promote a successful Build Docker image run into the Upload to DockerHub configuration or pipeline to re-deploy that artifact without rebuilding it. Learn more: Partial chain execution https://www.jetbrains.com/help/teamcity/2026.2/run-build-chains.html Partial+chain+execution Debug jobs You can now debug individual jobs without leaving Edit mode and triggering the entire pipeline. This small addition allows you to design complex workflows faster because you can fix problems as they arise instead of having to work your way through the failure log at the very end. Learn more: Debug jobs https://www.jetbrains.com/help/teamcity/2026.2/create-and-edit-pipelines.html Debug+Jobs Unbound pipelines You can now create pipelines without VCS roots attached. Previously, this option was only available for build configurations. Choose the Without repository option when creating a pipeline to create a custom workflow that does not check out any remote sources. Learn more: Create and edit pipelines https://www.jetbrains.com/help/teamcity/create-and-edit-pipelines.html New MCP tools Your AI Agents can now use three more tools designed to work with TeamCity pipelines: teamcity pipeline get , teamcity pipeline post , and teamcity pipeline delete . To avoid any accidents, the post and delete tools are only available in Brave mode https://www.jetbrains.com/help/teamcity/2026.2/ai-agent-integration.html Safety+Concerns , which has to be enabled manually. Learn more: TeamCity MCP https://www.jetbrains.com/help/teamcity/2026.2/ai-agent-integration.html TeamCity+MCP TeamCity AI AI continues to reshape software development. The TeamCity team is taking a practical approach: using AI to solve real CI/CD problems and remove friction from everyday tasks – all while keeping you in control of which AI capabilities you use and how you use them. AI Assistant AI Assistant now supports Bring Your Own Key BYOK , enabling you to plug in any supported LLM. Go to Admin | AI Assistant and enter your developer AI API key. Then you’ll be able to use Anthropic, OpenAI, Google Gemini, or any other supported model to inspect the builds and debug your CI workflows. Learn more: AI providers https://www.jetbrains.com/help/teamcity/2026.2/ai-assistant.html Providers MCP enhancements In version 2026.1, we have implemented an MCP endpoint that allows AI agents to use a TeamCity server as a source of tools. In this release, we’re adding three more tools that let you use your preferred AI solutions to manage pipelines in addition to classic build configurations and projects. Moreover, we have implemented the OAuth/PKCE authorization flow, which means you no longer need to manually issue a personal access token and pass it to your AI environment configuration. Learn more: TeamCity MCP https://www.jetbrains.com/help/teamcity/2026.2/ai-agent-integration.html TeamCity+MCP Rerun failed chain builds The Dependencies tab of build configuration settings now includes a Retry Settings group. Enable the Use custom retry settings option and enter a positive number of retry attempts in the corresponding box when you want to delay a downstream build and automatically retry a failing dependency in place – no need to rerun the entire chain. Learn more: Dependency retry settings https://www.jetbrains.com/help/teamcity/2026.2/configuring-dependencies.html Dependency+retry+settings IDE integrations We’ve updated our IDEA plugin https://plugins.jetbrains.com/plugin/25142-teamcity-experimental- so that it now comes with our CLI tool https://www.jetbrains.com/help/teamcity/teamcity-cli.html and offers seamless MCP server registration. This means you can set up integration with Junie or third-party AI agents with little to no manual setup. The plugin core functionality has also been improved: You can now view the results of TeamCity tests from the IDE and launch remote runs with local changes that have not yet been committed. In addition, Kotlin Multiplatform https://kotlinlang.org/docs/multiplatform/get-started.html developers working with IDEA and Android Studio can now install this plugin to claim an exclusive free-tier TeamCity Cloud subscription currently capped at 30,000 build credits per month where it’s possible to run automatically generated pipelines that test and build your iOS applications. The plugin also simplifies TestFlight integration, making the process of deploying your iOS applications significantly easier. GitHub integration The Pull Requests build feature can now match pull requests by their source branch instead of their branch reference. This lets TeamCity recognize separate pull requests in different repositories as related changes and build them together in the same build chain, even when GitHub assigns them different pull request numbers. Learn more: GitHub pull requests https://www.jetbrains.com/help/teamcity/2026.2/pull-requests.html GitHub+Pull+Requests DSL compilation mode For projects that store Kotlin DSL versioned settings in remote repositories, you can now choose where TeamCity runs this code: on the server or build agents https://www.jetbrains.com/help/teamcity/2026.2/kotlin-dsl.html DSL+Compilation . Executing kDSL on build agents is less restricting and more secure, but both options come with certain trade-offs, so make sure to check our documentation for all recommendations and caveats. Subscribe to TeamCity Blog updates