Junie (JetBrains)
Junie is JetBrains' LLM-agnostic AI coding agent that ships code from your terminal, IDE, or CI/CD pipeline. Give it a task in natural language, fix a bug, implement a feature, or review a PR, and it handles the rest. Rechecked 2026-08-21: the first-party repo is github.com/JetBrains/junie with 397 stars, created 2025-04-07. It is positioned as "your real coding buddy" for developers already in the JetBrains ecosystem.
Unlike single-vendor coding agents, Junie lets you choose the model behind it: authenticate with a JetBrains account, a Junie API key, or bring your own key (BYOK) for Anthropic, OpenAI, Google, xAI, OpenRouter, or Copilot.
Key Features
- Terminal-first agent: run
junieand describe a task, then let it build a project, analyze a branch, or fix failing tests. - IDE and CI/CD: integrates with JetBrains IDEs and ships code from CI/CD pipelines, not just the terminal.
- Bring Your Own Key: use your own provider (Anthropic, OpenAI, Google, xAI, OpenRouter, Copilot) or a Junie API key.
- GitHub automation: run
/install-github-actionso Junie responds to issues, PRs, and CI failures automatically. - Update channels: run stable, EAP, nightly, or experimental builds with
junie --channel=....
Use Cases
- JetBrains users who want a coding agent that works across terminal, IDE, and CI without leaving their toolchain.
- Model-agnostic developers who want to switch LLM providers rather than be locked to one vendor.
- Teams that want agents responding to GitHub issues, PRs, and CI automatically.
Pricing
| Plan | Price | Model |
|---|---|---|
| BYOK | Your key's cost | Use your own provider (Anthropic, OpenAI, Google, xAI, OpenRouter, Copilot) |
| JetBrains account | Per JetBrains AI plan | JetBrains-managed models |
| Junie API key | Per Junie tokens | Junie-provided models |
Pricing follows JetBrains' AI service plans and your chosen provider. Recheck current pricing at junie.jetbrains.com.
Advantages
- LLM-agnostic: choose your model and bring your own key, unlike more single-vendor agents.
- Multi-surface: terminal, IDE, and CI/CD in one agent.
- GitHub-native: automated responses to issues, PRs, and CI failures.
Getting Started
- Install:
curl -fsSL https://junie.jetbrains.com/install.sh | bash(macOS/Linux) ornpm install -g @jetbrains/junie. - Run
junieand authenticate (JetBrains account, API key, or BYOK). - Give it a task, such as "Fix the failing tests in the auth module and add a regression test."
- For GitHub automation, run
/install-github-actioninside the agent.
Frequently Asked Questions
Is Junie free?
The repo and CLI are open on GitHub; using models through BYOK costs what your provider charges. Recheck current pricing.
Which models can I use?
Anthropic, OpenAI, Google, xAI, OpenRouter, and Copilot are supported through BYOK.
Does it work without the JetBrains IDE?
Yes. Junie is terminal-first and works in CI/CD, so it does not require an IDE.
Alternatives
- Cursor: an AI-first IDE and agent platform.
- Claude Code: Anthropic's terminal coding agent.
- Codex CLI: OpenAI's coding agent in the terminal.
Tips
- Use
junie --eapto try the latest build once without changing your default channel. - Set up
/install-github-actionfor hands-off issue and PR handling. - Bring your own key to control model choice and cost.
Conclusion
If you live in the JetBrains ecosystem and want one LLM-agnostic agent that works in the terminal, IDE, and CI/CD, Junie is a strong fit. Start at junie.jetbrains.com or github.com/JetBrains/junie.
Comments
No comments yet. Be the first to comment!
Related Tools
Command Code
commandcode.ai
AI coding agent from Command Code AI that learns your coding taste with taste-1, then ships, reviews, and refactors in your style from the terminal.
OpenCode
opencode.ai
MIT open-source coding agent for terminal, IDE, and desktop. 198k GitHub stars. Bring any model, or use free included models and Zen.
Continue
continue.dev
Open-source coding agent for VS Code and JetBrains. Apache-2.0, local models, CLI, and MCP. Acquired by Cursor; the repo stays public.
Related Insights
Skills + Hooks + Plugins: How Anthropic Redefined AI Coding Tool Extensibility
An in-depth analysis of Claude Code's trinity architecture of Skills, Hooks, and Plugins. Explore why this design is more advanced than GitHub Copilot and Cursor, and how it redefines AI coding tool extensibility through open standards.
What Locks Codex Is the Picker, Not the Models
You already pay for OpenCode Go, Grok, and Z.ai, but the Codex picker still shows mostly GPT. The community project codex-router does not teach another install ritual. It puts subscriptions you already bought back into the selector, keeps keys on the machine, and leaves native GPT alone.
Hook OpenCode Go into Codex on Windows. Do Not Open a Second Toolkit.
A ChatGPT-signed Codex desktop app still shows mostly GPT in the picker. On Windows, enable only OpenCode Go and the Grok, GLM, Kimi, DeepSeek, and MiniMax models you already pay for appear in the same selector. Keys stay local. Native GPT stays put.