JetBrains Builds Centralized AI Cost Control Without Restricting Developer Tools

JetBrains faced a familiar challenge for fast-growing tech companies: AI spending spiraled out of control in just six months, roughly multiplying tenfold. Rather than locking down tool choices, the company constructed a shared infrastructure layer to track, measure, and govern AI use across the organization while preserving developer autonomy.

The Problem: Hidden Growth#

The explosion in spending caught up with reality when the company analyzed consumption patterns. Engineers were regularly juggling three to five different AI tools each month. Token usage spiked sharply starting in January 2026 as next-generation models like Claude Opus 4.5 and 4.6 became available, making it nearly impossible to track where money was actually flowing.

Initially, finance teams spent four days manually collating spending data from spreadsheets—a snapshot that became stale the moment it was finished. Understanding the actual bill across multiple providers and platforms proved unwieldy and reactive.

Visibility to Control#

The company first invested in automation, layering provider APIs and internal dashboards to give teams a continuous, real-time picture of spending and projected costs. Visibility alone, however, wasn't enough: dashboards could show where the problem was, but they couldn't intervene in the actual requests going out to AI providers.

The turning point came from an internal tool a single developer had already built. JetBrains expanded this prototype into a system called Central CLI—a unified command-line interface for accessing both its own AI products and third-party tools. By routing all requests through this common gateway instead of having them go directly to external providers, the company gained a central control point.

A Different Path Forward#

The approach stands apart from how other organizations have responded. Accenture told staff to reduce unnecessary AI use, while Uber introduced caps after burning through its annual budget in four months. JetBrains chose not to restrict the tool palette to one or two products, arguing that the market moves too fast to lock in choices for even a few months.