JetBrains Launches Air to Integrate AI Agents into IDE Workflows
New IDE feature enables developers to manage, review, and control AI coding agents directly.
1Managing the Agentic Bottleneck
As AI agents generate more code, developers face increasing pressure to review and validate automated outputs. Reports indicate nearly half of developers spend four or more hours weekly correcting AI prompts and code. JetBrains introduced Air in the 2026.3 release to centralize agent management, ensuring developers maintain control over their evolving codebases.
2Deep IDE Integration
Air functions as a workspace within JetBrains IDEs, allowing agents to leverage native intelligence. Through the IDE's Model Context Protocol server, agents can execute build configurations, query databases, and perform refactorings. This integration enables agents to utilize the same debugging and profiling tools that human developers use for verification and maintenance.
3Parallel Session Management
Developers can track multiple agent sessions simultaneously across various projects within a single interface. Air utilizes local Git worktrees to prevent collisions between parallel tasks, allowing users to review changes systematically. Notifications alert developers when an agent requires input, ensuring that automated workflows remain transparent and aligned with project requirements.
4Flexible Agent Compatibility
The platform supports a wide range of third-party agents, including Claude, GitHub Copilot, and Codex. Users can connect existing subscriptions or API keys directly into the IDE environment. For those without specific subscriptions, JetBrains provides the Junie agent in a free Lite configuration, accessible simply by signing into a JetBrains account.
5Review and Feedback Loops
Air includes a dedicated review panel where developers can inspect agent changes before final integration. Users can leave specific Code Notes on lines to request iterations or ask questions, maintaining context throughout the process. This mechanism ensures that developers retain ownership of the code while benefiting from the speed of automation.
6Transitioning from AI Assistant
Air replaces the previous AI Assistant chat, converting existing conversations into new agent sessions. The interface offers improved visibility into token usage and quota limits directly within the IDE. While AI Assistant remains supported until early 2027, Air provides a more robust, full-screen tab experience for complex terminal-based agent interactions.
7Developer-Centric Control
The architecture prioritizes developer autonomy, allowing users to choose exactly when and how to delegate tasks. Agents do not automatically start or transmit code without explicit authorization from the user. Developers can disable Air entirely via settings, ensuring the tool remains an optional enhancement rather than an intrusive requirement for daily coding.
8Future of Agentic Development
By embedding agentic workflows into the IDE, JetBrains aims to solve the maintenance challenges inherent in AI-generated code. This approach treats agents as collaborative tools rather than autonomous black boxes. Future updates plan to track agent-created worktrees, further streamlining the transition from AI-assisted prototyping to production-ready software development within professional environments.