JetBrains Junie

JetBrains Junie is a coding agent for JetBrains IDEs, command-line work and CI/CD tasks. It is a strong fit for developers who want to plan and supervise agent-led changes within JetBrains tools. Choose it for that control and workflow range; weigh its credit limits and the Windows sandbox gap before committing.

Overview

Junie combines chat assistance and agent mode with codebase indexing and terminal access. It can run in supported JetBrains IDEs, from a CLI on Linux, macOS and Windows, or headlessly in CI/CD scripts. Remote Control adds asynchronous CLI execution monitored through a web app. The range suits developers moving between local and automated work, but Windows users do not get CLI sandbox mode.

Key features

Planning and approvals

Advanced Plan Mode divides proposed work into requirements, design and delivery stages before editing code. Plans can be changed and committed in .junie/plans, which gives teams a durable review point before implementation. Junie also allows users to approve key actions; its CLI asks before sensitive actions unless brave mode is enabled or an action is allowed. These safeguards suit developers delegating substantial changes who still want to retain control.

Models and project context

Junie supports models from OpenAI, Anthropic, Google and xAI. The CLI accepts provider API keys directly, while Junie Lite / BYOK supports any AI model, including locally running models, with usage billed by the provider at its rates and no markup. Persistent project guidelines and Agent Skills packages help carry project-specific instructions into work. Slash commands are shared between CLI and IDE through ACP.

Remote and issue-driven work

Remote Control lets users run CLI tasks asynchronously and monitor them in a web app. GitHub Actions can start Junie from issues and pull requests, while GitLab CI/CD can run AI tasks from issues or merge requests. These integrations make Junie relevant beyond interactive coding, though they also make its value depend on whether a team wants an agent in those workflows.

Sandbox scope

The CLI sandbox is available on macOS and Linux, not Windows. It wraps terminal commands, but not Junie's file edits, MCP servers, hooks or git operations. That is a narrower boundary than the word “sandbox” might suggest; teams needing those activities covered should factor in the limitation rather than treating the mode as comprehensive isolation.

Pricing

Junie is freemium, with a 30-day trial. Its paid plans are billed per user per month, and the important distinction is the number of included AI credits, not simply access to the agent.

PlanPriceWhat it includesFree to Startfree5 AI credits; billed AI usage billed separatelyAI Pro$8.33 a month10 AI Credits per 30 days; anytime top-upsAI Ultimate$25 a month35 AI Credits per 30 days; anytime top-upsJunie Lite / BYOKfreeFree model included; provider usage billed separately, with zero markup; any AI model, including local modelsFree to Start is a way to try the workflow, but five credits and separately billed AI usage make it a limited free option. AI Pro is the lower-cost paid choice for individuals with modest, bounded use; its ten credits reset every 30 days, and top-ups are available when they run out. AI Ultimate raises that allowance to 35 credits for users expecting more activity, but it still has a quota. Both paid tiers are per-user subscriptions, so teams should account for seats as well as usage. Junie Lite / BYOK avoids a subscription charge and offers model flexibility, but shifts usage costs to the provider. No separate seat count is stated for the free plans.

Platforms

The IDE plugin works in IntelliJ IDEA Ultimate and Community, PyCharm Pro, WebStorm, GoLand, PhpStorm, RubyMine, RustRover, CLion, Rider and Android Studio; IDE version support varies by product. The CLI runs on Linux, macOS and Windows, and Junie also supports web-based monitoring for Remote Control. The platform breadth is useful for JetBrains users across operating systems, with the significant qualification that Windows lacks CLI sandbox mode.

Who it's for

Junie is best suited to developers already working in supported JetBrains IDEs who want an agent to plan, edit and use the terminal while leaving approval points in the workflow. Its CLI, CI/CD integrations and remote monitoring also suit teams that want to automate or supervise work beyond the editor. Developers who need a Windows CLI sandbox, or who want predictable costs without monitoring credits or provider usage, should look elsewhere.

Pros and cons

Pros: Planning produces editable, commit-ready stages, making agent proposals reviewable before code changes.

Pros: Multiple model providers and direct CLI API keys give users flexibility in how they supply models.

Pros: IDE, CLI, headless CI/CD and asynchronous remote workflows cover more than editor chat alone.

Cons: Included credits are limited even on paid tiers, so heavier use can require top-ups.

Cons: Free-plan AI usage can be billed separately, and BYOK shifts costs to the model provider.

Cons: Windows users cannot use CLI sandbox mode, and the sandbox on supported systems does not cover several important agent operations.

Alternatives

GitHub Copilot is worth comparing for readers who want a freemium coding assistant across a broad set of platforms, including Android and iOS. Kilo Code is another freemium option, with VS Code, JetBrains and CLI extensions. Claude Code is a paid alternative with a free plan and support across desktop and mobile platforms. Devin offers a freemium alternative with API access as well as web and desktop platforms. Tencent CodeBuddy and Tongyi Lingma are freemium alternatives with extension support on Linux, macOS and Windows. Atlassian Rovo Dev is another freemium choice, with extension and web support. OpenCode is a freemium alternative with API, extension, web and desktop platform support. Readers comparing categories can also browse AI Coding Assistants and AI Code Generators.

Verdict

Choose Junie if you work in JetBrains tools and want a coding agent whose plans, approvals and model options support more controlled delegation across IDE, CLI and CI/CD work. Its strongest reason to choose is that breadth paired with explicit review points. Look elsewhere if Windows sandboxing is essential or if credit quotas and separately billed AI use would make costs hard to manage.