JetBrains Air Governance extends cost and visibility tracking across AI providers

The Agent Client Protocol is where the design decision sits. JetBrains says ACP standardizes the connection between an IDE and an agent's full harness, including its planning, logic, tools, model routing, and observability ^[5]. Model routing therefore sits on the agent's side of that boundary. The IDE is a client. For a dollar figure to reach a cost view, the harness has to report which model it called and what the call cost, and the post does not list the usage fields that cross the boundary.

Practical coverage of a cross-vendor cost report is then the set of ACP agents that actually report usage. JetBrains says developers can discover and run a growing range of compatible agents through the ACP Registry while continuing to work inside JetBrains IDEs ^[6]. A pilot can measure that directly: run two agents from two vendors for a week, then check whether both appear in Air Governance with spend attached.

Air Governance is JetBrains Central under a new name ^[2]. Central arrived alongside the company's first public experiments with agentic development environments, six months before this post ^[12]. The post also dates its own progress "Since March" ^[15]. That puts the governance product at roughly six months old ^[16]. In between, JetBrains rolled out a Central CLI, shared context, cloud agents, automations, governance, and AI cost controls for teams and organizations ^[13]. Renaming a control plane is cheaper than building one, and this one was built first.

"The era in which the whole software development system can be contained in one window is ending," JetBrains wrote ^[10]. The same post says the IDE remains important to the company's future, and that the foundational agentic experience is being brought into JetBrains IDEs ^[11].

On who chooses the tools, the post argues against a handover. Developers "can choose the agent, model, or service suited to the task without forcing the organization to give up context, visibility, or control," JetBrains wrote ^[8]. Selection stays with the developer. The reporting and the policy move up a level. The company gives as a core belief that no single model, agent, or service will be right for every developer, team, or task ^[17].

For a team already on JetBrains IDEs, adopting this costs a second product to configure and a named owner for its policies. Junie, JetBrains' coding agent for professional software development, will be supported across all Air surfaces ^[4], which makes the all-JetBrains path the shortest one to a working pilot.