Air expands beyond a single agent workspace
JetBrains on 22 September brought its recent agent-development work under the Air name, defining a system that spans agent orchestration inside JetBrains IDEs, shared team workflows and organisation-level governance. The company says Air in JetBrains IDEs is the developer-facing surface for directing agents and reviewing their changes, Air Teams coordinates cloud projects and automations, and Air Governance is the renamed JetBrains Central control layer for policy, auditability, cost management and provider access.
The current product page presents Air as multi-vendor rather than tied to JetBrains' own Junie agent. It lists Claude Agent, OpenAI Codex, Junie, GitHub Copilot and OpenCode, alongside other agents that implement the Agent Client Protocol. ACP gives the IDE a common connection to an agent's broader harness, including planning, tools and model routing, while JetBrains keeps code review and deterministic IDE analysis in the same workflow.
Availability differs across the three layers
The IDE experience can be tried now, including by individual developers, and JetBrains documents local use with provider subscriptions, API keys or compatible custom endpoints. Air Teams is labelled early access and is aimed at organisations that want browser-based projects, cloud execution, shared automations, managed secrets and centrally configured MCP connections. JetBrains therefore has a shipping local surface and a less mature shared-cloud surface under the same Air umbrella.
Those deployment paths do not have identical provider freedom. JetBrains documentation says local desktop work can use bring-your-own-key or provider accounts, while cloud tasks use JetBrains AI credits and remain subject to organisation policy. Gemini CLI is also documented as temporarily unavailable for cloud tasks. The multi-vendor claim is established for the system, but individual agents and authentication methods still have surface-specific constraints.
Governance is moving into the development stack
Air Governance extends the former Central product around rules, permissions, audit history, cost controls and visibility across providers. JetBrains is pairing that with Air Teams controls over execution environments and external connections, so the system is designed to let organisations approve different agents without giving up a shared policy and accounting layer. The company also describes Air Context as a semantic indexing service intended to give agents structured repository knowledge across these surfaces.
That architecture targets a practical problem created by mixed agent fleets: code can be produced through several IDE, terminal and cloud tools while review, spend and access rules remain fragmented. JetBrains' argument that its code intelligence can make agents cheaper or more accurate is still a company claim. The announcement establishes the product structure and control surfaces, but independent evidence on reliability, security boundaries, cost savings and review quality is still limited.
Community attention is evidence of interest, not adoption
A Hacker News discussion helped surface the 22 September announcement, but comments there are not used to establish the product's capabilities or uptake. The Confirmed status comes from JetBrains' announcement and current product material, which directly establish the Air naming, supported agent ecosystem, governance layer and early-access state of Air Teams. No community reaction is treated as evidence that developers have migrated to Air or that the system performs better than competing agent environments.
JetBrains also describes future mobile and remote experiences, richer shared context and more work triggered by repository events, schedules and delivery processes. Those items remain direction rather than shipped capability until corresponding releases appear. The material change today is narrower: JetBrains has turned Air into an umbrella for local agent work, cloud team coordination and cross-vendor governance, with different parts of that system at different stages of availability.