Cloud tasks
JetBrains Air Teams tasks run in containerized cloud environments instead of on your machine. Run a task in the cloud when you want an isolated, reproducible environment, want to work from any device, or want to automate workflows that do not require direct user involvement.
To quickly run your first task in the cloud, proceed to Quickstart.
Where you can start a task
To run a task, you must be signed in with your JetBrains Account.
In the browser – open https://air.jetbrains.cloud and run tasks there. See Work with a task in the browser.
In an IntelliJ-based IDE – use AI Assistant in an IDE such as IntelliJ IDEA, PyCharm, or WebStorm. See Run cloud tasks from IntelliJ-based IDEs.
In the Air app – start a task and select Cloud as the run environment.
When to run in the cloud
Running in the cloud is useful when local execution is inconvenient or not possible.
You do not have the project on your machine – you can open the project remotely and work with it without cloning or setting it up locally.
You work from different places or devices – you can start and manage tasks from the browser, not only from the Air app.
You need automations – scheduled and event-based workflows run in cloud environments and do not depend on your machine being online. Learn about automations.
You want stronger isolation – tasks run in a remote environment and do not use your local files, local processes, or local credentials unless you explicitly provide the required configuration.
You want reproducible execution – cloud environments can be configured per repository, so tasks run with the same base image, variables, and setup regardless of who starts them.
How this differs from running locally
Running in a cloud environment differs from running on your machine in several ways.
The environment is remote – the task runs in a cloud environment, not on your machine. Unlike a local run, the environment does not already have your repository checked out. Before you can run cloud tasks, make the repository available to JetBrains Air Teams. Learn how to connect repositories.
The repository is cloned shallow – the environment gets only the latest commit, so the Git history isn't there until someone fetches it. See Repository checkout and Git history.
The environment does not inherit your local setup automatically – if your project needs specific dependencies, environment variables, secrets, or other setup, you must provide them as part of the cloud configuration. If you do not create one, JetBrains Air Teams uses a default environment, which is enough for simple tasks. Learn how to configure cloud environments.
The environment is suspended when idle – after 75 minutes of inactivity, JetBrains Air Teams suspends the environment and marks the task Suspended. You can resume it until JetBrains Air Teams archives the task after 24 hours of inactivity. See Cloud environment lifecycle.
Results are committed for you – JetBrains Air Teams commits what the agent produced to a separate task branch and pushes it after every agent turn, so the changes are not lost when the environment is no longer available.
The review and apply workflow is different – there is nothing to accept. You review what already landed, iterate on it, and create a pull request from the task branch. Learn more in What happens to your changes.
Cloud environment lifecycle
JetBrains Air Teams creates a cloud environment when a task starts and suspends it after a period of inactivity. A cloud task shows one of these states:
State | What it means |
|---|---|
Starting | JetBrains Air Teams is creating and starting the cloud environment for the task. |
Done | The task has finished. The environment is still available, so you can review the result and continue working in it. |
Suspended | JetBrains Air Teams stopped the environment after 75 minutes of inactivity. Your results are safe on the task branch. Click Resume to restart the environment and continue working on the task. See Resume a suspended environment. |
Archived | JetBrains Air Teams archived the task after 24 hours of inactivity. You can still open the task to view its chat history, but you can't resume it, because the environment can no longer be restarted. Your results remain on the task branch. |
Problem | The task ran into an error. See Troubleshoot cloud tasks. |
Snapshots and preserved state
A cloud environment keeps its state on disk. When a run finishes, JetBrains Air Teams can save that state as a snapshot and start a later run from it instead of preparing a new environment from scratch. This makes the run start faster, and it means a run doesn't always begin in a clean container.
How much of the earlier state a run gets depends on whether you resume the same task or start a new one.
You resume the same task – the environment returns as it was, including uncommitted working-tree changes and everything outside the repository, such as installed packages, build caches, and ~/.local. Running processes are the exception, so restart servers, watchers, and anything else the task had running.
You start a new task on the same repository – the snapshot provides the prepared environment: the image, installed dependencies, build caches, and other files outside the repository. The repository is always synced fresh – JetBrains Air Teams discards leftover working-tree changes and checks out the branch that the new task requests.
Snapshots are a startup optimization, so there's nothing to configure. Tasks and automation runs on the same repository share a snapshot, and snapshots are personal – yours is never reused for someone else. Snapshots also expire, so a run can still start in a cold environment and take longer.
Treat preserved state as a speed-up, not a guarantee. If a run needs a file or a tool, install it from the startup script rather than relying on an earlier run leaving it in place. What the script installs is saved in the snapshot, so an expensive download usually happens only once.
Credits and usage
Cloud tasks always spend JetBrains AI credits from your subscription. A provider account of your own – a subscription or a BYOK key you connected in the Air app – works for local tasks only, so in the cloud JetBrains Air Teams runs every agent on the JetBrains AI provider.
Cloud tasks draw on the same monthly credit limit as your local ones. To see how much of it you've used, see Check how much you've used.
Automation runs are cloud tasks too. A personal automation spends your credits, and a project automation spends the project's – see AI credits for automation runs.
If you belong to an organization, an administrator sets your credit limit in JetBrains Central Console under . The same administrator controls whether you can run cloud tasks at all and which agents are available there. See Manage AI access for your organization.