JetBrains announced Air Teams, a team-focused layer within the Air ecosystem for coordinating coding agent work and sharing what works across the software development lifecycle. The service is currently available to JetBrains business customers, while the company plans to expand access later to include individual customers.
The service targets a practical problem in using AI agents within development teams: each developer’s settings, instructions, and environment often remain on their computer, making it difficult to transfer a successful workflow to the rest of the team. Air Teams therefore moves these elements into a shared space that the team can manage and reuse.
Four components for a shared workflow
Air Teams consists of four interconnected elements:
- Automations: Recurring tasks that agents execute in the cloud, such as reviewing pull requests, handling specific issues, or updating dependencies, and that begin based on an event or schedule.
- Shared cloud environments: Configurations that include the tools, dependencies, and access data required for building and testing, allowing the team to configure them once and reuse them.
- Cloud tasks: Work that can run in parallel without keeping the developer’s computer occupied, with the ability to start and monitor it from the integrated development environment or browser, and from a phone later.
- Team projects: A space that brings together members, environments, connectors, and automations, with shared roles and credits.
Automations control repetitive work
Each automation is built from four components: instructions that define what the agent is required to do, a runtime environment, tools it can use, and a trigger that determines when it starts. Triggers include events from GitHub or Jira, webhooks, and schedules, with support for other types in the future. The service also supports connectors to tools such as Jira, Figma, and Linear.
JetBrains presents examples that include reviewing every new pull request, attempting to fix small issues in YouTrack and opening a pull request for review, and updating dependencies twice a week. In the last example, the agent builds the project and runs the tests, then fixes failures or reverts the update if it cannot be fixed, while maintaining one updated pull request instead of flooding the team with repeated requests.
What changes in practice?
Air Teams turns agent configuration into a shared asset rather than an individual experiment. The repository environment is stored in the .air/cloud/startup.sh file inside the repository, and it can be reviewed and versioned like the rest of the code. An agent can also help configure it by inspecting the repository and performing the installation and build, then proposing the script in a separate branch for review.
JetBrains says that sharing an environment does not mean sharing the credentials themselves; shared secrets can be used without exposing their values, while personal secrets and repository permissions remain tied to each user. Projects grant automations a service account and their own AI credits, allowing them to continue running even after their creator leaves the team, if administrators choose to use project credits.
Keeping the decision in the engineer’s hands
The platform attempts to reduce the noise caused by unnecessary comments and pull requests by keeping changes in the form of pull requests that the engineer decides to merge or reject. Each run also retains the agent conversation log and tool calls, allowing the team to inspect the reason for the result when an error occurs. However, the source provides no details about pricing, usage limits, or broader governance mechanisms, points that teams will need to verify before adopting the service at production scale.