JetBrains has revealed its strategic response to a tenfold surge in internal AI spending: rather than restricting developer tools, the company built a centralized proxy and accounting layer called Central CLI to monitor and govern LLM consumption in real-time.
- JetBrains' development-related AI expenses increased tenfold in six months due to multi-tool usage and advanced frontier models.
- Instead of restricting tools, JetBrains built 'Central CLI' to route all AI traffic through a centralized, managed proxy.
- This centralized proxy allows managers to apply spending caps and use internal credits on third-party LLM APIs.
- Over 1,000 developers adopted the platform, showcasing how engineering teams can reconcile FinOps goals with tool flexibility.
The cost of prototyping: jetbrains' 10x AI surge<\/h2>
AI adoption among software development teams has evolved at a blistering pace, but so has the bill. JetBrains recently revealed that its internal generative AI expenses skyrocketed tenfold over a six-month period. This cost explosion was primarily fueled by developers leveraging three to five different AI companions concurrently, alongside a sharp rise in token utilization starting in early 2026 as teams began querying high-end frontier models like Claude Opus 4.5 and 4.6.
Initially, tracking down these costs was a highly manual, tedious process. Finance and engineering managers spent days scraping provider APIs and manually compiling usage spreadsheets just to get a delayed snapshot of the organization's LLM footprint. Realizing this pattern was unsustainable, JetBrains shifted its strategy from passive tracking to proactive infrastructure management.
# publicidade
Central CLI: merging choice with financial control<\/h2>
Rather than following the industry trend of banning certain external models or forcing developers onto a single approved platform, JetBrains took an engineering-first approach. The company adapted an internal, developer-created tool wrapper into what is now 'Central CLI.' This tool serves as an standardized local gateway for invoking both proprietary and third-party AI models.
By routing CLI-based AI requests directly through JetBrains' pre-existing internal AI platform, the infrastructure team effectively established a middleman in the API traffic flow. This architecture enables the platform to participate in the actual request generation cycle rather than simply auditing bills after they arrive. Under this new model, managers can set specific spending ceilings for individuals, departments, or projects, while applying the company's internal AI-credit tokenomics directly to external providers.
Adoption, roadblocks, and the wider finops landscape<\/h2>
The rollout of Central CLI was met with rapid organic adoption, with over 1,000 internal developers integrating it into their workflows within weeks. Despite the success, JetBrains acknowledges that some gaps remain. Hard-to-track terminal agents and individual developer subscriptions occasionally bypass the central gateway, and creating a perfectly fair budgeting algorithm across diverse product teams remains a work in progress.
JetBrains' infrastructure-centric solution places it at the forefront of a growing movement. Organizations globally are struggling with the financial realities of generative AI. The FinOps Foundation has already expanded its scope to include AI cost tracking, and other enterprises have taken more aggressive steps—such as Uber introducing hard caps after consuming an annual AI budget in a mere four months, or Accenture issuing direct guidelines to restrict non-essential queries. JetBrains, by contrast, proves that structural proxies can curb waste while letting developers build with the best tools available.
| Company | Cost Management Strategy | Developer Impact | Primary Mechanism |
|---|---|---|---|
| JetBrains | Centralized API Proxy (Central CLI) | High flexibility (retains tool choice) | Real-time routing, quotas & shared credits |
| Uber | Strict monthly quotas | Medium flexibility | Hard limits after exhausting annual budget in 4 months |
| Accenture | Behavioral curbs | Low/Medium flexibility | Advising employees to limit unnecessary usage |
Arraste para o lado para ver toda a tabela.
Perguntas frequentes (faq)<\/h2>
Why did jetbrains' AI costs increase so quickly?<\/h3>
JetBrains saw its development-related AI spend grow tenfold in six months due to developers utilizing three to five distinct tools monthly, combined with a sharp increase in token volume triggered by expensive frontier models like Claude 4.5 and 4.6.
What is central CLI?<\/h3>
Central CLI is an internal wrapper developed by JetBrains that acts as a proxy, routing third-party and internal LLM requests through the company's existing AI platform for centralized monitoring, credit allocation, and budget enforcement.
How does jetbrains' strategy differ from other tech companies?<\/h3>
While companies like Uber and Accenture implemented strict limits or manual restrictions to curb spend, JetBrains chose an infrastructure-driven approach that preserves developer tool choice while routing usage through a monitored proxy layer.
