JetBrains spent years building AI pair programming tooling. Now it is retiring that product and replacing it with something structurally different: not a better AI, but a governance layer for autonomous agents. The architect of the pair programming era is betting the bottleneck has moved.
image from grok
JetBrains announced it will retire Code With Me, its AI pair programmer IDE product, and replace it with JetBrains Central, an agent-agnostic governance platform for autonomous AI agents that treats AI as an operator rather than a coding assistant. The company is positioning Central's core value around enterprise observability, cost attribution, and policy enforcement for agent-driven development, arguing that organizations are repeating the same ROI miscalculation that plagued early cloud adoption. The platform supports external agents including Claude, Codex, and Gemini CLI alongside JetBrains tooling, signaling a shift from building exclusive AI features to selling infrastructure for an increasingly agent-driven development ecosystem.
JetBrains built its reputation on a simple premise: AI as a pair programmer, always on, inside the IDE. On March 24, 2026, the company announced it would retire Code With Me, the product that embodied that premise, and replace it with something structurally different. Code With Me's final supported version drops in 2026.1; the relay infrastructure goes dark in Q1 2027. Simultaneously, JetBrains Central enters early access, an open governance platform for agentic development that treats AI not as a pair programmer but as an autonomous operator. The pivot is the story.
Oleg Koverznev, JetBrains' VP and head of Agentic Platform, put the economic logic plainly in a briefing with The New Stack: the industry is about to replay the cloud ROI crisis. The parallel is apt. In the early cloud era, enterprises bought compute before they understood what governance actually cost. Now they're buying AI agent tooling on the same terms, and the bill is arriving faster than expected.
JetBrains' own data illustrates the gap between adoption and integration. Ninety percent of 11,000 developers surveyed in January 2026 already use AI at work, but only 13 percent use it across the full software development lifecycle. Twenty-two percent already use AI coding agents, with another 66 percent planning to adopt within 12 months. The trajectory is clear; the organizational infrastructure to manage it is not. What JetBrains is really selling with Central is observability and cost attribution, the ability to see what agents are doing, enforce policies, and understand where money goes when autonomous code generation runs at scale.
Central has three core capabilities: governance and control, agent execution infrastructure, and agent optimization and context. The governance layer is where the real product lives. It handles policy enforcement, identity and access management, observability, auditability, and cost attribution for agent-driven work, according to InfoWorld. Crucially, Central is not exclusive: it supports external agents including Claude Agent, Codex, Gemini CLI, and custom-built solutions alongside JetBrains' own tooling. This is not a walled garden. It is infrastructure designed to be agent-agnostic, which is itself a bet on how the market will consolidate.
Code With Me was collaborative editing, two humans or a human and an AI working in the same IDE session. Agents operating autonomously across an enterprise codebase need something different. The correction tax data suggests why. Nearly 30 percent of senior engineers say fixing AI output consumed most of the time AI saved them, compared to 17 percent of junior developers, citing a Fastly survey. As code generation becomes cheaper, the cost of managing what comes out of it goes up.
The broader ROI picture is not kind to the industry. Forbes reported that fewer than 10 percent of enterprises report measurable ROI from AI investments, with MIT finding a 95 percent failure rate for enterprise generative AI projects not showing measurable financial return, as covered by CIO. Amazon's internal review documents originally cited AI-assisted changes as a factor in a trend of operational incidents before the language was later revised, views confirmed by both CNBC and the Financial Times. Toby Ord, a senior researcher at the Oxford Martin AI Governance Initiative, has suggested current benchmarks overstate what AI systems can do in real development environments. The numbers are directional, not definitive, but the pattern is consistent: the tooling is ahead of the management layer.
JetBrains is not publishing data from Code With Me itself, the platform it ran for years with millions of developer sessions measuring whether AI pair programming actually worked at scale. That absence is notable. Hadi Hariri, JetBrains' SVP of Operations, said the company is piloting Central internally as a governance and observability response to agentic workflows, with the early access program launching in Q2 2026 with a limited group of design partners. The company has operational data on AI-assisted coding at scale and is choosing to ship a governance platform rather than publish findings. What they found is not for public consumption.
The irony is not lost on the industry. JetBrains spent years convincing developers that AI pair programming was the future of the IDE. Now the company that built that era is building the infrastructure to manage what comes after it. Koverznev's framing is the thesis of a different product category. Code generation is cheap and no longer a bottleneck; the challenge is managing operational and economic complexity. Central is JetBrains' answer to that challenge.
Whether it lands depends on whether enterprises actually deploy it. The early access program is small by design. Central supports external agents, which is the right architectural call. Agents built by Anthropic, Google, and OpenAI will not run exclusively inside JetBrains tooling. The governance layer has to span the ecosystem, or it solves a narrower problem than the company claims. The question is not whether agentic development needs governance. It is whether JetBrains is the right company to build it, and whether enterprises will pay for it before the next wave of tooling arrives.
What JetBrains has done is signal where the bottleneck has moved. From code generation to code governance. From IDE collaboration to organizational observability. The cloud ROI crisis happened because compute became cheap before anyone built the tools to track it. AI agent tooling is on the same trajectory. JetBrains is betting that the next essential layer is not a better model. It is a better ledger.
Story entered the newsroom
Research completed — 0 sources registered. JetBrains announced Central (March 24, 2026) — a governance and execution platform for AI coding agents. VP/Head of Agentic Platform Oleg Koverznev wa
Draft (852 words)
Approved for publication
Published (890 words)
Get the best frontier systems analysis delivered weekly. No spam, no fluff.
Agentics · 2h 47m ago · 3 min read
Agentics · 5h 34m ago · 4 min read
