JetBrains Air and Junie CLI: Multi-agent coding without the overhead
JetBrains announced two tools on March 9 that aim to make AI a practical part of daily development: Air, an environment for running multiple coding agents side by side, and Junie CLI, an LLM-agnostic coding agent you can use across your stack.
If you've been juggling terminals, browser tabs, and context windows just to compare model output, this is a cleaner path. One workspace, multiple agents, and code changes shown in the context of your repo.
Air: An agentic development environment for concurrent agents
Air is in public preview and free on macOS, with Linux and Windows coming soon. Download it at air.dev.
Think of Air as a single workspace where agents like Claude Agent, Gemini CLI, Codex, and Junie CLI can run concurrently. You can switch between them without losing context, compare outputs, and keep everything anchored to your codebase.
Tasks can reference exact context-line numbers, commits, classes, methods, or symbols-so you avoid pasting giant blobs of text. When an agent finishes, Air shows diffs in the context of the entire repo, alongside essentials like a terminal, Git tools, and a built-in preview.
More agents are coming via the Agent Client Protocol (ACP) through the ACP Agent Registry, so teams can standardize on one environment while keeping flexibility on the models they use.
- Availability: Public preview, free on macOS
- Core value: Run and compare multiple agents at once with repo-level context
- Tooling: Integrated terminal, Git, and preview for faster review and iteration
Junie CLI: LLM-agnostic coding agent, built for real workflows
Junie CLI is in beta and focused on being model-neutral and grounded in your codebase by default. Grab it at junie.jetbrains.com.
It supports top models from OpenAI, Anthropic, Google, and Grok, and plans to integrate new releases as they land. The goal: solve complex problems, respect context, and be reliable and secure across environments.
- Where it runs: Directly in the terminal, inside any IDE, in CI/CD, and on GitHub or GitLab
- Model strategy: LLM-agnostic with a bias for high-performing options
- Release timing: Planned March release for broad workflow support
Practical ways to use Air and Junie CLI together
- Compare agents on the same task: Use Air to run Claude, Gemini, Codex, and Junie side by side. Keep the best diff and discard the rest.
- Targeted tasks with precise scope: Point agents to a function, class, or commit. Keep changes local and reviewable.
- CI assistance without noise: Run Junie CLI in CI to propose patches, test fixes, or comment on PRs. Review results in Air before merging.
- Model fit per task: Pick the agent that handles refactors well for code changes and another that's better at writing tests or docs.
Getting started fast
- Install Air on macOS and add your preferred agents in one workspace.
- Use Junie CLI locally from the terminal, then wire it into your IDE and CI.
- Scope tasks with code symbols instead of pasting large context. Review diffs in Air and commit with Git tools on the spot.
Why this matters for engineering teams
This setup reduces context thrash and makes agent output measurable. You can run head-to-head trials across models, standardize review, and keep everything anchored to your repo, not a chat window.
If your team is testing LLMs for refactoring, test generation, or bug-fix suggestions, Air and Junie CLI give you a practical, auditable path to production use.
Learn more
Your membership also unlocks:
