Junie

Junie is what happens when the company that built the IDE builds the AI agent inside it. Unlike bolt-on assistants that treat the editor as a text buffer, Junie uses JetBrains' own inspections engine, refactoring APIs, test runner, and VCS integration.

The result is an agent that does not simulate development, it uses the same tools you use. In Code mode, Junie breaks tasks into multi-step plans, writes code across files, runs the project's actual test suite, and loops back to fix failures until tests pass.

That test-verification loop, where Junie writes code, runs tests, discovers bugs, and iterates automatically, is the feature most consistently praised in practitioner reviews.

The March 2026 launch of Junie CLI in beta fundamentally changed the tool's trajectory. Previously locked to JetBrains IDEs, Junie CLI is a standalone, LLM-agnostic agent that runs in any terminal, any IDE, and CI/CD pipelines. With BYOK (Bring Your Own Key) support for OpenAI, Anthropic, Google, and Grok models, teams can choose their preferred LLM without a JetBrains AI subscription.

This moves Junie from a JetBrains-only perk to a genuine Claude Code and Codex CLI competitor. Real-time prompting and next-task prediction round out the CLI experience, though the beta still lacks the deep IDE integration that makes the in-editor version compelling.

Pricing is structured around JetBrains AI plans: $10/month (AI Pro, 10 credits/30 days), $30/month (AI Ultimate, 35 credits), and $60/month (AI Enterprise, maximum credits), where 1 credit equals $1.00 of compute. BYOK usage requires no subscription at all, making Junie CLI effectively free for teams with their own API keys.

The main limitation remains ecosystem breadth, in-IDE Junie still requires a JetBrains IDE, and while the CLI version is editor-agnostic, it is in beta and lacks the semantic project understanding that makes the IDE version special. Teams fully invested in VS Code should evaluate whether the CLI beta is mature enough before committing.