JetBrains Air Governance: Selling the Fix for AI Cost Sprawl

JetBrains has reorganized its agentic development tooling into a family of products under the Air name. The most revealing piece is Air Governance, a renamed version of what the company previously called JetBrains Central 1. Air Governance is a cost and policy layer for AI-assisted development, and it grew out of a problem JetBrains first faced inside its own engineering organization.

What JetBrains announced

JetBrains describes its strategy as a shift from a single product to an "open system of products" for agent-driven software development 1. The lineup has three parts:

• Air in JetBrains IDEs lets developers direct and orchestrate agents and check their output inside the IDE.
• Air Teams coordinates and automates delivery workflows that involve both humans and autonomous agents.
• Air Governance covers organizational policy, visibility, auditability, cost management and accountability 1.

The company says it started experimenting publicly with agentic environments about six months ago. Over that period it introduced Central and then added a CLI, shared context, cloud agents, automations and AI cost controls 1.

Air Governance is meant to work across providers and tools. Developers can pick whichever agent, model or service suits a task, and the organization still keeps visibility and spending oversight 1. Under this design, governance does not require standardizing on one vendor.

The internal cost problem behind it

According to TechGig, JetBrains' AI development spending grew tenfold in six months 2. The company built a centralized system around the Central CLI to give it visibility and spending limits without restricting which tools developers use 2. TechGig reports that more than 1,000 JetBrains developers adopted the CLI within weeks 2.

A post on Mastodon adds that JetBrains' own blog conceded the company "didn't know how to control them systematically" before it introduced cost controls 4. The post calls this a rare admission that puts a number on the AI spending inside a developer-tools company 4. Vendors rarely publish cost curves like this. A tenfold increase in two quarters shows how quickly agentic workflows can outgrow informal budgeting once developers start running agents routinely.

The sources broadly agree on the facts but frame them differently. JetBrains' blog presents Air Governance as a natural step in building a product system 1. TechGig and the Mastodon post start from the cost spike 24. Both framings are accurate. In practice, JetBrains used its own engineering organization as a test case and is now selling the result.

Why governance is becoming a product category

The timing matters because coding agents are spreading across many surfaces at once. The Mastodon post lists Claude Code, Codex CLI and Gemini CLI as the major terminal agents and names Warp as a fourth contender 4.

Warp shows what heavy agent use looks like at scale. According to OpenAI, Warp open-sourced its terminal client with OpenAI as founding sponsor of the repository, and introduced a model it calls Open Agentic Development. In that model, humans set objectives and supervise, while agents plan, write code, test and open pull requests 3. OpenAI says agents now co-create about 90% of Warp's pull requests 3. It also says that running long agent workflows taught Warp that scaling them requires observability, coordination, memory and human review 3.

The user figures differ by source. OpenAI says Warp has nearly 1 million developers and is used by more than 56% of the Fortune 500 3. The Mastodon post says Warp claims more than 1 million active developers, and dates the open-sourcing of the terminal and its Oz cloud-agent orchestration platform to April 2026 4. The difference is small, but readers should treat these as company-reported numbers.

As more agents from more vendors enter the workflow, it becomes harder to track spending in one place. Warp's list of needs, including observability and coordination, overlaps heavily with what Air Governance offers 13. Companies that build agent tools are reaching similar conclusions about what is missing.

The takeaway

JetBrains' main bet is not on any particular agent. It is betting that organizations will use many agents and will need a neutral layer to monitor and limit them. Its cross-provider design reflects a view that developers will not accept being locked into one model or tool 12.

The approach has a credibility advantage, since JetBrains says it ran into the problem itself before productizing a fix 24. It also faces a test. A governance layer is only useful if it actually captures spending across the third-party agents, terminals and cloud services that teams adopt. With terminal-native tools like Warp gaining ground 34, keeping visibility across all of those tools may be harder than building the dashboard.

Even so, the tenfold figure is likely to resonate with engineering leaders. It suggests that the next round of competition in AI developer tooling will include cost control and accountability alongside raw agent capability.