JetBrains Air wants to make AI coding agents answer to the whole team

JetBrains has introduced JetBrains Air, a platform for agentic software development that brings models, coding agents, team workflows and organisational controls into one system. The important shift is not another chatbot in an IDE, but an attempt to manage AI-assisted coding across an entire development operation.

Watch Desk analysis

What happened

In its 22 September announcement, JetBrains describes Air as an open system spanning individual IDEs, coordinated team workflows and organisational governance. The platform is designed to let developers use different models and agents while preserving shared context, visibility into costs, security controls and policy management.

JetBrains also says Air combines those agentic features with its existing deterministic code intelligence. In plain English, the company is trying to put probabilistic AI tools inside a development environment that still understands the structure and rules of software, rather than leaving teams to bolt together a model, a coding agent and a compliance spreadsheet.

Why it matters

The practical problem is familiar to any team that has tried to scale AI coding beyond enthusiastic individuals. One developer chooses a model, another chooses an agent, a third invents a permissions scheme, and finance eventually discovers that “just trying it” has acquired a bill. Air’s pitch is that these choices can be coordinated without forcing everyone onto a single model vendor.

That multi-vendor approach could matter as coding agents become more capable and more deeply embedded in repositories, build systems and production workflows. Teams may want to switch models as prices, performance or data policies change. A shared control layer could make that less painful, though JetBrains has not supplied independent evidence in the announcement showing how well Air performs in real-world deployments.

Our read

This is a serious infrastructure play, not merely an IDE feature wearing an agent-shaped hat. The strongest idea is the attempt to connect individual developer tools with team coordination and governance. That is where many AI coding experiments become difficult: the code may be impressive, but nobody is quite sure which model changed it, what it cost or whether it followed the organisation’s rules.

The announcement establishes JetBrains’ product direction, not that Air has solved those problems. Developers should look for working documentation, model and agent compatibility, permission boundaries and evidence from teams using it at scale before treating the platform as a cure for coding-agent sprawl.

What to watch

• Model choice:
  whether developers can switch providers without losing context or useful workflow features.
• Governance:
  how Air handles permissions, sensitive code, audit trails and policy enforcement.
• Cost visibility:
  whether teams can understand and control spending across models and agents.
• Developer evidence:
  whether early users report fewer integration headaches rather than merely prettier demos.

Discussion spark: Should teams standardise on one coding-agent platform for control, or preserve a multi-vendor setup even if it makes governance harder?

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.