JetBrains Central: When AI Agents Become Too Many for Manual Control

Until recently, AI in development was perceived merely as an assistant within the code editor. It suggested line completions, offered autocomplete, and occasionally explained someone else's code. This was useful, but rather modest.

Today, the picture has changed. AI agents no longer just provide hints – they independently investigate tasks, generate code, run tests, and execute multi-step scenarios. Moreover, they do this not just in one editor window, but simultaneously across multiple environments, tools, and build pipelines.

According to the January JetBrains AI Pulse survey, which involved 11,000 developers worldwide, 90% of respondents already use AI in their work. Agents – that is, systems acting more autonomously – are used by 22% of those surveyed, while 66% of companies plan to implement them within the next 12 months.

But there is a problem: despite the technology's popularity, the real impact is still limited. No more than 13% of developers use AI throughout the entire software development life cycle (SDLC) – for example, during code reviews or the release process. Most organizations cannot transform AI usage into measurable improvements: neither delivery speed, system reliability, nor development costs are changing significantly.

Challenges of managing autonomous AI agents in development

The problem isn't the agents; it's managing them

When there is only one agent, everything is relatively clear. But when there are dozens, each working in its own tool, in its own environment, and by its own rules – chaos ensues. It becomes unclear who is doing what, how much it costs, whether the result can be trusted, and how it all fits into the team's actual processes.

This is exactly the challenge JetBrains set out to solve by introducing the JetBrains Central platform.

In short: this isn't just another AI assistant or a new code editor. It is an agentic development management system – a «control center» of sorts that connects tools, agents, and infrastructure into a single whole.

Key features and capabilities of JetBrains Central

What's inside

JetBrains Central is built around three core capabilities.

First is governance and control. The platform handles access policies, permissions, action audits, and cost accounting. Simply put, an organization can see what agents are doing, who started them, and how much it costs.

Second is infrastructure for running agents. Agents can operate in the cloud, securely and independently of a specific developer's machine.

Third is shared context and task routing. Agents gain access to a semantic layer – accumulated knowledge about the code, architecture, and system behavior in production. This allows them to work not «blindly», but with an understanding of how a specific system is built. Based on this context, the platform also directs tasks to the most suitable models and tools.

Integration with third-party AI agents and IDEs

Openness as a principle

An important point: JetBrains Central is designed as an open system without being locked into a specific set of tools or services. Developers can run agentic workflows from JetBrains IDEs, third-party editors, the command line, or a web interface. Supported agents include both JetBrains' own solutions and external ones – such as Claude Agent, Codex, Gemini CLI, or proprietary corporate developments.

This approach means that organizations don't need to abandon their established processes. JetBrains Central integrates into existing infrastructure rather than requiring its replacement.

Collaborative workflows between developers and AI agents

Agents and humans in the same flow

A particular emphasis in the platform is placed on human-AI collaboration. Agents interact with teams through familiar tools: Slack, Atlassian products, and Linear. The idea is for workflows to remain transparent and integrated into the team's actual activities, rather than existing in isolation.

To support this, JetBrains is also developing Air Team – a space for coordinating tasks between employees and agents within a team. It is built on top of JetBrains Central and allows for organizing tasks, running multi-step processes, and maintaining transparency over what is happening.

"We are increasingly using agents and AI processes, which creates a need for better cost visibility and management. This is why we have started piloting JetBrains Central internally. It is an evolutionary process that reflects how we build products at JetBrains: using our own tools to better understand and shape them."

This is how Hadi Hariri, Senior VP of Operations at JetBrains, commented on the launch.

Early Access Program and target audience for the platform

When and for whom

An Early Access Program is planned for the second quarter of 2026. For now, it is open to a limited circle of design partners who will be able to test the platform in real-world conditions.

The product's positioning is quite broad. Individual developers gain the freedom to choose tools while maintaining control over the process. Teams get the ability to coordinate the work of humans and agents in a shared context. Organizations get centralized management of costs, access rights, auditing, and AI scaling.

The impact of agentic development on software engineering

What this means in practice

Looking at the bigger picture, JetBrains is betting that agentic development is not a passing trend, but a structural shift. And in this new paradigm, what's needed is not just a set of «smart» features, but a system that makes their work predictable, manageable, and economically transparent.

For now, it looks like a statement of intent – the platform has not yet been fully launched. However, the direction is clear: JetBrains wants to occupy the assembly point where agents, tools, teams, and business objectives intersect.

Time and the first results from Early Access partners will tell how successful this solution proves to be. 🔍