Air

Air is a specialized desktop application developed by JetBrains, designed as an agentic development environment for managing AI-assisted coding workflows. Unlike single-agent chat interfaces, Air orchestrates multiple AI coding agents to work concurrently on different tasks within a single project. It integrates leading AI agents including Claude Agent, Codex, Gemini CLI, and Junie, allowing developers to match the right tool to each job. Air provides structured workflows for defining tasks with precise code context, running agents in isolated environments to prevent conflicts, and reviewing changes with full code intelligence.

How to use Air

Air operates as a standalone desktop application, primarily for macOS with previews for Windows and Linux coming soon. To use Air, developers first define a coding task by exploring their project and gathering context through references to files, commits, symbols, or images. They then select and launch one or more supported AI agents (Codex, Claude Agent, Gemini CLI, or Junie) to execute the task. Air handles the setup using isolated execution methods like Docker containers or Git worktrees to run tasks in parallel without interference. Users can monitor all active agent progress from a central overview, quickly switch context to provide additional input, and finally review, accept/reject changes, and commit the results back to their local project.

Core Features of Air

• Concurrent Multi-Agent Execution: Run multiple AI coding agents (Claude Agent, Codex, Gemini CLI, Junie) side-by-side on different tasks without conflicts.
• Isolated Task Environments: Automatically sets up isolated execution using Docker, Git worktrees, or (coming soon) cloud environments for each agent task.
• Code-Aware Task Definition: Define tasks with precise context by referencing specific files, code symbols, commits, or images from the project.
• Centralized Progress Oversight: Get an overview of all agents' progress on async tasks, enabling easy monitoring and course correction.
• Integrated Review Workflow: Review agent-generated changes at the file level, accept or reject code comments, and prepare changes for commit within Air.
• Agent-Agnostic Flexibility: Switch between or compare outputs from different AI agents (Claude, OpenAI, Google, JetBrains) for the same task.
• Workspace Management: Use workspaces to scope agent sessions, Git state, and tools to a single project, preserving history and avoiding branch collisions.

Use Cases for Air

Air is designed for software developers and teams engaged in AI-augmented coding. Key use cases include: rapidly prototyping or building features in unfamiliar languages or frameworks by delegating to specialized agents; breaking down large development tasks into smaller sub-tasks executed concurrently by different agents; experimenting with and comparing outputs from multiple AI coding models (Claude, GPT, Gemini) on the same problem; and maintaining oversight and control when automating repetitive coding or refactoring workflows with AI assistance.

Advantages of Air

Air's primary advantage is shifting focus from managing individual AI models to orchestrating a complete, controlled development workflow. It increases velocity by enabling true parallel task execution rather than sequential agent use. It provides better delegation by allowing developers to break work apart and assign the most suitable agent to each sub-task. Enhanced oversight and visibility let developers monitor progress and intervene when needed without losing context. Finally, its integrated, code-intelligent review process improves the quality and integration of AI-generated code into the main project.