Application development and project management platform provider JetBrains is on a mission to arm developers with a broad set of artificial intelligence (AI) functions spanning non-technical and hardcore coding tasks.
In June last year, the Czech company released Grazie, an AI writing companion for “people working in tech” designed to streamline non-coding tasks by integrating GenAI into JetBrains’ integrated development environments (IDEs), browsers and other tools. Grazie’s non-coding AI capabilities focused on tasks like documentation, offering instant proofreading, inline text completion, summarization, translation and rephrasing.
This month, the company has come forward with Junie, an AI coding agent, that offers agentic power for code co-creation – once again, right inside the IDE. Andrew Zakonov, product leader for JetBrains says that software engineers using Junie can “fully delegate routine tasks” to a personal coding agent or collaborate with it.
Junie in A Bottle
By combining the JetBrains IDE with what the company promises are “reliable LLMs”, Junie has already shown itself capable of solving a variety of developer tasks that typically take some hours of work. Junie can solve 53.6% of tasks on a single run, according to SWEBench Verified, a curated benchmark of 500 developer tasks. The technology sits on a waitlist early access program at the time of writing.
“Such a success rate is promising – this shows the potential to also adapt Junie to the reality of today’s software development, including large amounts of tasks of different complexity. Junie will unlock the power of coding agents for millions of developers and companies around the world,” claimed an upbeat Zakonov.
Currently engineered for Python, Kotlin and Java use (with Mac OS X and Linux platforms currently supported), Junie’s main USPs, Zakonov details, are tasks related to fixing bugs and resolving issues related to code compilation. Possibly dovetailing with Grazie, Junie can also help with updating project documentation and provide an option to add new code tests while also helping with problem resolution related to existing tests that have recorded failures.
“Getting started with Junie is as simple as installing it into your IDE. You can then begin with delegating simple tasks as you get used to working with the coding agent, so you don’t need to make changes to your workflow,” said Zakonov.
When software engineers are more comfortable, they can integrate Junie into a team workflow and start redefining tasks to boost productivity and “unleash ingenuity” with coding experiences powered by agentic AI, the company says.
Software engineers using Junie will be able to review proposed changes, maintain project context and guide critical decisions so that they stay in control, even when delegating tasks. They can always review code changes and decide how the agent executes commands.
“[In reality], AI-generated code can be just as flawed as developer-written code. Ultimately, Junie will not just speed up development – it is poised to raise the bar for code quality. By combining the power of JetBrains IDEs with LLMs, Junie can generate code, run inspections, write tests and verify they have passed,” detailed Zakonov, in a JetBrains blog.
Contextual Code Comprehension
Junie is designed to understand the context of any given project, so it can adapt to the users’ unique coding style. The agent can also follow specific coding guidelines, which, in theory, helps enhance its ability to align with the software developers’ codebase.
“This results in better code quality and control on how Junie performs tasks, ensuring reliability, making Junie a trusted collaborator on your team,” said Zakonov. “At JetBrains, we build products together with users, listening to the feedback to drive innovation in software development. This approach helps us create products empowering millions of developers to make anything happen – with code.”
Junie is currently available in the JetBrains IDEs – IntelliJ IDEA Ultimate and PyCharm Professional with WebStorm coming next.
