JetBrains released Air Teams, a shared cloud layer for running coding agents so the whole team can reuse what one developer sets up

An agent that finishes work on one developer's laptop still sits idle everywhere else on the team, because the setup that made it work is only on that laptop. JetBrains introduced Air Teams on 28 September to move that setup off the laptop and into a shared workspace, so a colleague and an Automation can run the same agent in the same environment on the same terms. The product is available today to JetBrains business customers, with individual access planned later, and is written by Vladimir Gromozdin on the Air blog.

Air Teams comes in four parts, and each fills one gap the post names. Automations handle recurring work such as code review, issue fixes and dependency updates, triggered by an event or a schedule instead of a person. Shared cloud environments hold the tools, dependencies and credentials the agent needs, set up once by one person and reused by every task on the team. Cloud tasks run those environments in parallel without using anyone's laptop while they run. Team projects tie the three together with their own service account and their own credits, so an Automation keeps running after its creator leaves the company.

The ten Automation templates ship with the release

The post shows a screenshot naming ten Automation templates included at launch, covering code review, bug fixes, dependency upgrades and documentation maintenance. Each combines four things (instructions, environment, tools and trigger), configured once and then reused on every run. JetBrains describes three it runs on its own team: a reviewer that reads a new pull request, posts inline comments and updates its review on each new commit; an issue fixer that opens a pull request when a small ticket in YouTrack is tagged; and a twice-weekly dependency updater that upgrades the project, runs the tests, fixes what breaks or reverts what cannot be fixed, and closes an older pull request before opening the new one.

The setup lives in the repository

An environment in Air Teams is defined by a .air/cloud/startup.sh script that lives in the same repository as the code, so the team versions and reviews it the same way as every other change. A team member picks the VM size, decides which domains the machine can reach, and adds the variables and secrets the build needs. An agent can also do the first pass: it inspects the repository, runs the install and build, asks for missing secrets, and commits a startup script to a branch for a human to review. Once the environment is ready, Air saves it as a snapshot and shares it. Shared secrets let a teammate or an Automation read a value without seeing it; personal secrets and repository access stay with each person.

The trade cloud agents make

The post includes a side-by-side comparison the reader can act on before switching. A local agent starts against the existing working copy, uses the developer's tools, and shares the machine's CPU and memory with everything else. A cloud agent starts the environment and copies the repository first, uses the tools and credentials the team configured, keeps working after the developer closes the laptop, and shares one environment across the team's tasks. The cost of the local path is that each teammate still prepares their own machine; the cost of the cloud path is a longer first start, unless the environment has been saved as a snapshot with cached state.

Source

• Primary: Introducing Air Teams, JetBrains Air blog