Delv
CodingActive· 6dby JetBrains★ 4.3

Junie

Autonomous coding agent by JetBrains that ships code from the IDE, CLI or CI/CD, powered by any LLM you choose.

Try it Source
A
Safety & Trust

Delv Safety Grade: A

Score 82/100 · assessed 2026-04-18

Maintainer95
Permissions55
Supply chain85
Transparency90
Incidents100

Junie is JetBrains' autonomous coding agent that operates within their IDE ecosystem, CLI, and CI/CD pipelines. As a major established vendor with decades of IDE development experience, JetBrains provides strong organisational legitimacy. The agent is open source on GitHub with clear documentation and active development. However, as an autonomous coding agent, Junie requires substantial permissions: filesystem write access to modify codebases, shell execution for build and test operations, repository write access for commits, and network access for LLM communication. The 'autonomous' nature means it can make changes without per-action approval, which inherently carries risk despite JetBrains' reputation. Supply chain is solid through standard distribution channels (JetBrains Toolbox, npm for CLI). The ability to use any LLM adds flexibility but also means security depends partly on your chosen model provider. No known security incidents. Best suited for teams already invested in JetBrains tooling who understand the trade-offs of autonomous agents.

Green flags

  • Maintained by JetBrains, a major established IDE vendor since 2000
  • Open source on GitHub with active development and clear documentation
  • Integrates with existing JetBrains security and permission systems
  • No known security incidents or CVEs
  • Supports self-hosted LLMs for data sovereignty

Red flags

  • Autonomous agent can make filesystem and repo changes without per-action approval
  • Shell execution capability required for build/test operations
  • Security depends partly on user's choice of external LLM provider
  • Paid product bundled with AI Pro limits accessibility for security review

Permissions requested

Read filesWrite filesShell executeOutbound networkRepo readRepo writeExternal LLM call
Assessed by Delv Editorial using public metadata. Grades are advisory and update as the ecosystem changes. They do not replace your own review of permissions and code before granting an agent access to sensitive systems.

Pricing

PAID — Bundled with AI Pro plans

Platforms

jetbrainscli

Review

Junie sits inside your JetBrains IDE and does what you'd normally ask Cursor or Copilot to do, except it keeps going without you. Point it at a bug or feature request, and it reads your codebase, writes tests, iterates on failures, and opens a PR. The autonomy is real: I've watched it spend twenty minutes refactoring a service layer while I reviewed another branch. The multi-step planning works better than expected. It doesn't just generate code and stop; it runs tests, reads error messages, adjusts, and tries again. Where it shines: medium-complexity tasks with clear acceptance criteria. I gave it a GitHub issue asking for pagination on an API endpoint. It added the query params, updated the repository layer, wrote integration tests, and caught a bug in the offset logic before I'd even looked. That iterative loop is the difference between this and a standard autocomplete tool. You're not babysitting every step. Failure modes are predictable. On a gnarly legacy codebase with inconsistent patterns, it thrashed for a while before I killed the run. It also struggles when the task is vague. "Make the checkout flow better" produces mediocre guesses. You need specificity: "Add a discount code field to the checkout form and validate it against the promotions table." The CLI and CI/CD modes are useful for teams that want agents in their pipeline, but the IDE integration is where it feels native. Compared to Cursor's Composer or Windsurf's Cascade, Junie has tighter JetBrains integration and better test-running loops. Cursor is faster for quick edits; Junie is better when you want to walk away. The bundled pricing with AI Pro means you're paying for the whole JetBrains AI suite, which is fine if you already live in IntelliJ or PyCharm, less appealing if you're a VS Code holdout. I'd reach for this when I have a well-defined task, a test suite that runs quickly, and thirty minutes to let it work. It's not replacing senior judgement, but it's a capable junior who doesn't need constant supervision.
Verdict

If you're already on JetBrains IDEs and want an agent that iterates without handholding, Junie earns its spot. VS Code users should stick with Cursor or Windsurf unless they're willing to switch ecosystems.

Good at

  • Actually iterates on test failures without human intervention
  • Native JetBrains integration feels polished, not bolted on
  • CLI and CI/CD modes let you run agents in pipelines
  • Works with any LLM, not locked to one provider
  • Handles medium-complexity tasks end-to-end reliably

Watch out

  • Bundled pricing means you pay for the full AI Pro suite
  • Struggles with vague or exploratory tasks
  • Less useful on legacy codebases with inconsistent patterns
  • Not available for VS Code users
  • Slower than Cursor for quick, single-file edits

Use cases

  • bug fixing
  • feature implementation
  • PR review