JetBrains puts AI agent usage behind one console, turning AI budgets from reports into enforced limits
Synopsis
JetBrains has released Air Governance, a console that gathers the AI agents, models and providers a team uses, sets AI credit limits for individuals, teams and service accounts, and tracks both cost and output. Previously each tool carried its own contract, dashboard and admin settings, so a cross-vendor picture of usage was hard to assemble; JetBrains says its own AI spend grew roughly tenfold in the first half of 2026 and its first cross-vendor usage report took four days to compile by hand. The product is available to all JetBrains organizational customers on free and paid plans and works with Claude Agent, Codex, Gemini CLI, Junie and other ACP-compatible agents.
Interpretation
JetBrains released Air Governance as the organization layer of JetBrains Air, letting admins decide in one console which AI agents, models and providers teams may use and set AI credit limits for individual users, teams and service accounts. Previously each AI tool came with its own contract, dashboard and admin settings, so when teams used agents from several providers at once it was hard to see who used which agents and models, how much AI cost the company, and whether it helped. JetBrains uses itself as the reference: its AI spend grew roughly tenfold in the first half of 2026, its first cross-vendor usage report took four days to compile by hand, and the tools now manage the agentic usage of roughly 2,300 people at JetBrains every month.
The console makes budgets enforceable rather than merely reported, and lets an organization connect its own model accounts centrally, approve which agents are allowed, and configure MCP servers at company level. Admins can combine approved providers, models, agents and credit limits into AI policies for different user groups, and when a new agent or model appears it is added to the same policies without a new contract or separate admin setup. The product spans developer surfaces including IDEs, the CLI, web browsers and mobile, works with Claude Agent, Codex, Gemini CLI, Junie and other ACP-compatible agents, and supports JetBrains AI or a customer's own OpenAI, Anthropic, Amazon Bedrock or Google Vertex AI accounts.
Cost and output are viewed in one place, with AI costs broken down by user, by agent, and by model and provider, alongside outcome metrics such as the split between human-authored and AI-generated code. Explaining to leadership what AI spend delivers was harder than increasing it, and spend can now be attributed down to each developer and compared across Junie, Claude Agent, Codex, Gemini CLI and other agents. How each metric is calculated is described in separate AI analytics documentation, and the product is available to all organizations, with the Business plan at USD 20 per month per developer with AI access, no individual AI licenses, and a shared pool of AI credits the organization buys separately.
Perspective
The demonstrated scope is JetBrains organizational customers using the console to manage cross-vendor agent usage, with the product available to all organizational customers on free and paid plans, across IDEs, the CLI, browsers and mobile, with compatibility for ACP-compatible agents; JetBrains says later releases will work on propagating MCP servers to third-party CLIs, supporting GitHub Copilot and cloud agent connections through Air Gateway, advanced AI analytics, credit limits and custom pricing for BYOK models, and managing skills across teams.
The spend growth, reporting time and monthly user count describe JetBrains' own situation, and the material gives no comparison data from other organizations; the outcome metrics list only the human-authored versus AI-generated code split and the share of AI-generated or AI-modified code, without saying how those metrics should be read as changes in productivity.