JetBrains Air
JetBrains · JetBrains Air: Run, Coordinate, and Govern Claude Agent, Codex, Junie, and Other Coding Agents
JetBrains Air is a multi-vendor system for running coding agents such as Claude Agent, Codex, Junie, Gemini CLI, and other ACP agents, and for governing them across a company. It runs agents side by side in JetBrains IDEs and a desktop app, runs cloud tasks and event-triggered automations for teams, and adds policies, credit limits, and session analytics for organizations. It suits engineering teams that use several agents and want one place to run, review, and pay for them.
Best for
Engineering organizations on JetBrains IDEs whose developers use several coding agents and that want parallel agent sessions in the IDE, shared cloud tasks and automations, and one place to set agent policies, credit limits, and BYOK providers
Not ideal for
Developers who use one agent and do not need orchestration, VS Code or Neovim users who want agents inside their editor (Air's editor integration is JetBrains-only), teams that need production-stable tooling today, and organizations that require self-hosting or Bitbucket or Azure DevOps support
Who it's for
Engineering teams, platform leads, and engineering managers at companies that use JetBrains IDEs and several coding agents, plus individual JetBrains users who want to run agents in parallel
Capabilities
- Runs Claude Agent, Codex, Junie, Gemini CLI, GitHub Copilot, OpenCode, and other ACP-compatible agents, detecting local installs and adding more from the ACP agent registry
- Air in JetBrains IDEs: a free plugin with a structured chat for ACP agents and a full terminal tab for CLI agents, plus a Global Prompt that attaches selections, files, symbols, test results, VCS changes, console output, or screenshots
- Parallel sessions across projects and worktrees, with unread updates, changed files, outgoing commits, and cost tracked in one view, and sessions restored after an IDE restart
- Review agent diffs with the IDE's inspections and tools and leave line comments for the agent to act on; supported agents can call IDE tools over MCP
- Named presets that save an agent, model, reasoning effort, and access level for reuse
- Air app for macOS, Linux, and Windows that runs local tasks in your workspace, a Git worktree, or a Docker container, with one plan mode for every agent and a review pass that can use a different agent
- Air Teams cloud tasks: each runs in a Docker container on a JetBrains-hosted VM (Small 2 CPU/8 GB up to Extra Large 16 CPU/64 GB), commits to its own task branch, and keeps going when you close your laptop
- Automations triggered by a schedule, a fixed interval, GitHub pull request, issue, or Actions failure events, Jira issue events, or an incoming webhook, with built-in templates such as Code Review, Issue Triage, and Generate Release Notes
- Shared projects with members, roles, environments, variables and secrets, network access modes (off, specified domains, trusted domains, all), and automations that run on project credits under a service account
- MCP connectors for Confluence, Figma, GitHub, GitLab, Grafana, Jira, Linear, Sentry, TeamCity, and YouTrack, or a repository .mcp.json for internal servers
- Air Governance (formerly JetBrains Central): AI access for users, groups, and service accounts, monthly credit limits, and policies that allow or block agents, MCP servers, BYOK providers, and cloud tasks
- Organization BYOK for OpenAI, Anthropic, Google Vertex AI, and Amazon Bedrock, plus a session explorer showing who ran which agent and model, for how long, and at what token and credit cost
- Flexible AI access in the IDE: an agent's own login, your provider subscription or API key, or JetBrains AI credits billed at public API rates
Limitations
- Most of the system is pre-release: Air in IDEs is a public preview with roughly weekly updates, Air Teams is in early access, and Air Governance features are in an Early Access Program
- Cloud tasks and automations always spend JetBrains AI credits, whatever agent you pick; your own Claude, ChatGPT, or API-key accounts work only for local tasks
- A JetBrains AI subscription is currently required for cloud runs, and JetBrains says cloud runs reach remaining customers gradually over the coming months
- In the IDE plugin, a Claude subscription cannot authorize Claude Agent in Air's chat surface; you need JetBrains AI or Anthropic API billing, or you run Claude Code in the terminal view. Docker tasks in the Air app cannot use a Claude subscription, and JetBrains' Air app docs are inconsistent on whether a Claude Pro or Max plan works for other local tasks
- Agents you add through acp.json run only on your machine and are not available for cloud tasks, and Gemini CLI is temporarily unavailable for cloud tasks
- Air Teams is for organizations, and setup requires an administrator in JetBrains Central Console to enable it and grant repository access
- Supported VCS hosts are GitHub, GitHub Enterprise Server, GitLab, and GitLab Self-Managed; there is no documented Bitbucket or Azure DevOps support
- No self-hosted deployment is documented; hybrid deployments, AI audit logs, and IdP/SCIM configuration are listed as coming soon for the Enterprise governance plan
- The IDE plugin needs a recent IntelliJ-based IDE: it installs from JetBrains Marketplace in 2026.2 versions and is bundled in 2026.3 EAP builds, and it is not available in Android Studio
Use cases
- Running Codex and Claude Agent on two branches of the same project in the IDE and comparing their diffs before keeping one
- Setting up a Code Review automation that runs on every new pull request in a GitHub organization
- Triggering an agent when a GitHub Actions workflow fails so it investigates and opens a fix pull request
- Starting an agent from a Jira status change or a webhook from an internal tool, with the payload passed as task context
- Giving every developer access to approved agents and models with a monthly credit limit, and checking the session explorer for unusually expensive runs
- Routing an organization's agent usage through its own Amazon Bedrock or OpenAI account with BYOK
Our take
JetBrains Air is aimed at the organization rather than the individual developer. Its value is that it lets developers keep whichever agent suits a task while giving the company one place to see, limit, and pay for all of them, with JetBrains' IDE code intelligence for reviewing what the agents produce. For a JetBrains shop that already has developers on Claude Code, Codex, and Junie, Governance plus the free IDE plugin is the low-risk starting point. Air Teams' cloud tasks and automations are more ambitious and less proven, and they always bill through JetBrains AI credits, so model the credit spend before turning on automations across many repositories. Treat the whole system as a fast-moving preview for now, and wait if you need self-hosting or support beyond GitHub and GitLab.
Who should use it
Engineering organizations on IntelliJ-based IDEs that already use several coding agents and want shared policies, credit limits, and analytics; platform teams that want reusable cloud environments and automations for code review, triage, or CI failures on GitHub or GitLab; and JetBrains users who want to run multiple agents in parallel and review their diffs in the IDE.
Who should skip it
Developers committed to one agent who do not need orchestration or governance, teams on VS Code or other editors who want agent sessions inside their editor rather than in a separate desktop app or browser, organizations that need self-hosting or Bitbucket or Azure DevOps today, and teams that cannot accept preview-quality tooling in their workflow.
Strengths
- Vendor-neutral: runs agents from Anthropic, OpenAI, Google, GitHub, JetBrains, and any ACP agent side by side
- Agent diffs are reviewed with the IDE's code intelligence and tools
- Cloud tasks and automations cover the work individual agent tools leave out: schedules, GitHub, Jira, and webhook triggers
- Governance spans agents and providers: access, policies, BYOK, credit limits, and per-session cost
- The IDE plugin is free and can use your existing agent subscriptions
Weaknesses
- Most components are preview or early access and change frequently
- Cloud work always spends JetBrains AI credits rather than your own agent subscriptions
- In the IDE plugin, Claude subscriptions only work through Claude Code in the terminal view
- No self-hosting, and no Bitbucket or Azure DevOps support documented
JetBrains Air pricing
Air in IDEs and Air app
Free
- Free plugin from JetBrains Marketplace
- Use agent logins, your own provider subscriptions or API keys, or JetBrains AI
- Junie with its Lite configuration free for a limited time (terms apply)
Governance Basic
$0
- SSO and access management
- Centralized licensing and billing
- AI spending limits and AI analytics
Governance Business
$20/seat/month
Billed monthly
- AI policies
- External AI providers (BYOK) and model management
- ACP-compatible agents
- Analytics API
- Free preview for organizations that opt in to the Early Access Program
Governance Enterprise
Custom
- Everything in Business
- Hybrid deployments, AI audit logs, workspaces, IdP and SCIM, and priority support, listed as coming soon
Free tier limits: The IDE plugin is free, but agent usage is billed by whichever account the agent uses: the agent's own subscription, your provider account, or JetBrains AI credits. The Governance Basic plan is free; AI credits are not included in any plan.
Note: JetBrains AI usage is metered in AI credits, where LLM interaction costs 1 AI credit per $1 spent on the provider side at public API rates. Organizations buy credits separately; regular credits expire after 12 months, and credits included in licenses follow the license term. Cloud tasks and automations always spend JetBrains AI credits, and a JetBrains AI subscription is currently required for cloud runs. JetBrains says services in Early Access Preview, including cloud agent and automation compute and context engine queries, may be free of charge. The Business plan's per-seat price applies after JetBrains Central reaches general availability and is charged for every user with AI access.
Technical specs
Where JetBrains Air excels
Standardizing agent access across an engineering organization
Instead of each developer buying and configuring agents separately, administrators grant AI access by group, choose which agents and BYOK providers are allowed, and set monthly credit limits, then use the session explorer to see which agents and models are used and what each run costs.
Automating pull request review and CI failure follow-up
An Air Teams automation can start from a new pull request or a failed GitHub Actions workflow, run an agent in a shared cloud environment with the right toolchain and network rules, and either comment or open a fix pull request for a person to merge.
Comparing agents on the same task inside the IDE
Air in JetBrains IDEs can run Codex, Claude Agent, and Junie in parallel sessions on separate worktrees, so a developer can review each diff with the IDE's inspections and keep the best result.
JetBrains Air vs. competitors
JetBrains Air vs. OpenHands
OpenHands is MIT-licensed and self-hostable. It ships its own model-agnostic agent plus Agent Canvas for running Claude Code, Codex, and Gemini CLI in worktrees, and an automation service with cron, GitHub, and Slack triggers. JetBrains Air is proprietary and JetBrains-hosted for cloud work, but it pairs agents with JetBrains IDE code intelligence and adds organization governance, BYOK policies, and credit limits. OpenHands fits teams that want to own the stack, and Air fits JetBrains shops that want a managed system.
JetBrains Air vs. Replicas
Replicas runs third-party harnesses such as Claude Code, Codex, and Cursor in cloud Linux VMs, using your own model credentials, and bills for VM minutes. Air Teams also runs multiple vendors' agents in cloud environments with automations, but cloud work bills through JetBrains AI credits, and Air adds an IDE plugin, a desktop app, and organization-level governance. Replicas fits teams that want to keep their own agent subscriptions in the cloud, and Air fits teams that want IDE, cloud, and governance from one vendor.
JetBrains Air vs. Orca
Orca is a free, open-source desktop app for running CLI agents side by side in git worktrees, with your own agents and subscriptions. Air in JetBrains IDEs and the Air app do the same local job inside or next to a JetBrains IDE, and Air adds cloud tasks, team automations, and governance. Orca suits individuals who want a lightweight, open-source multi-agent workspace, and Air suits teams that need the organizational layer.
JetBrains Air vs. GitHub Copilot
GitHub Copilot spans many editors and runs a cloud coding agent on GitHub issues, with third-party agents such as Claude and Codex on higher tiers, all billed through GitHub plans. JetBrains Air is vendor-neutral across agents and providers, supports GitLab and self-managed GitLab as well as GitHub, and centers on JetBrains IDEs and JetBrains AI credits; Copilot's own agent is also one of the agents Air can run. Copilot fits GitHub-centric teams across editors, and Air fits JetBrains-centric teams that want governance across several agents.
Frequently asked questions
What is JetBrains Air?
JetBrains Air is JetBrains' system of products for building software with coding agents. Air in JetBrains IDEs and the Air app run agents such as Claude Agent, Codex, Junie, and Gemini CLI side by side. Air Teams runs agent tasks and automations in shared cloud environments. Air Governance (formerly JetBrains Central) gives organizations policies, credit limits, BYOK providers, and analytics for agent use.
How is JetBrains Air different from Junie?
Junie is JetBrains' own coding agent. Air is the layer that runs and governs agents, Junie included, alongside agents from other vendors such as Claude Agent, Codex, and Gemini CLI. JetBrains says Junie will be supported across all Air surfaces, but Air does not require you to use it.
How much does JetBrains Air cost?
The Air in IDEs plugin is free, and agents can use the accounts or subscriptions you already have. Agent use through JetBrains AI is paid in AI credits, where 1 credit equals $1 of provider API spend. Cloud tasks and automations in Air Teams always spend JetBrains AI credits, and a JetBrains AI subscription is currently required for cloud runs. For governance, JetBrains Central lists a free Basic plan, a Business plan at $20 per seat per month, and Enterprise on request; AI credits are bought separately. During the Early Access Program, organizations are on Basic and can preview Business features for free.
Which agents does JetBrains Air support?
Air lists Claude Agent, Codex, Junie, GitHub Copilot, OpenCode, and Gemini, and any ACP-compatible agent can be added from the ACP registry or an acp.json file. Cloud tasks in Air Teams are limited to the agents your organization enables, agents added through acp.json are not available in the cloud, and Gemini CLI is temporarily unavailable for cloud tasks.
Can I use my Claude or ChatGPT subscription with JetBrains Air?
For local work, yes, with limits. In the IDE, a Claude subscription works only when you run Claude Code in Air's terminal view; Claude Agent in the chat view needs JetBrains AI or Anthropic API billing. In the Air app, Codex can use a ChatGPT Plus, Pro, or Team account, while JetBrains' documentation is inconsistent on whether Claude Agent accepts a Claude Pro or Max plan or needs Anthropic API billing, and Docker tasks cannot use a Claude subscription. Cloud tasks and automations always use JetBrains AI credits instead.
What happened to JetBrains Central?
JetBrains Central, introduced in March 2026, became Air Governance when JetBrains announced JetBrains Air on September 22, 2026. Administrators still manage it in JetBrains Central Console, where they control AI access, credit limits, policies, agents, MCP servers, and BYOK providers.
Can JetBrains Air be self-hosted?
No self-hosted deployment is documented. Air Teams runs cloud tasks on JetBrains-hosted VMs, and hybrid deployments are listed as coming soon for the Enterprise governance plan. The Air app and IDE plugin can run agents locally on your machine.
Integrations & fit
Alternatives to consider
OpenHands
Developers and platform teams who want an open-source, bring-your-own-model coding agent they can run and automate on their own machines or VMs
Freemium
Replicas
Engineering teams already using Claude Code, Codex, or Cursor who want those agents running in configured cloud VMs, triggered from Slack, Linear, GitHub, GitLab, or automations, with PRs that follow up on CI failures and reviews
Paid
Stably AI
Developers who already use one or more CLI coding agents and want to run and review many of them in parallel without branch conflicts
FreeAbout JetBrains Air
JetBrains Air is not another coding agent. It is the layer JetBrains built to sit above agents, including ones it does not make, and JetBrains announced it on September 22, 2026 as a system of products. Air in JetBrains IDEs is a free plugin, currently a public preview, that turns an IntelliJ-based IDE into a workspace for agents. It detects agents already installed on your machine or adds them from the ACP registry. ACP agents get a structured chat, and CLI agents run in a full terminal inside an editor tab. You can keep sessions running across several projects and worktrees and review each agent's diff with the IDE's own tools. The Air app is a standalone desktop client for macOS, Linux, and Windows that runs tasks in your workspace, a Git worktree, or a Docker container. It also works with no organization at all if you bring your own provider account. Air Teams, in early access at air.jetbrains.cloud, moves work to the cloud. Each task runs in a Docker container on a JetBrains-hosted VM, built from a shared environment configuration, and its work goes to a task branch that a person merges through a pull request. Automations start agents on a schedule, on GitHub or Jira events, or from a webhook, and MCP connectors give agents access to Jira, Linear, Sentry, Figma, YouTrack, TeamCity, and other services. Air Governance, formerly JetBrains Central, runs in JetBrains Central Console. Administrators decide who gets AI access, which agents, MCP servers, and BYOK providers (OpenAI, Anthropic, Google Vertex AI, Amazon Bedrock) are allowed, and how many credits each person can spend each month. A session explorer shows every agent run with its tokens and cost. Junie, JetBrains' own coding agent (listed separately), is supported on every Air surface. The tradeoffs: most of the system is still preview or early access and changes weekly, cloud tasks and automations always spend JetBrains AI credits rather than your own Claude or ChatGPT plan, and in the IDE plugin a Claude subscription works only through Claude Code in the terminal view, not in Air's structured chat.
Updates from JetBrains Air
Plugin build 263.6259.32 added a dedicated Worktree settings panel, fixed Codex sign-in through ChatGPT OAuth and device-code flows, and added message actions for copying Markdown, forking, and reporting issues.
JetBrains Central Console, which hosts Air Governance, added Anthropic as an organization AI provider, GitLab Self-Managed connections, and a page showing how licenses were converted to a shared pool of AI credits.
JetBrains brought its agentic tools together as JetBrains Air: Air in JetBrains IDEs, Air Teams, and Air Governance (formerly JetBrains Central), with Junie supported across all Air surfaces and other agents connected through ACP.
JetBrains Central Console gained an audit page that records changes to users, roles, AI access, and AI credit limits, including who made them and whether they succeeded, failed, or were rejected.
A session explorer began showing every AI session in an organization, with the user, agents and models used, duration, tokens, and AI credits, for agents run through JetBrains IDEs, Air, and JetBrains Central CLI.
Are you the founder? Claim this listing →