You have 1 article left to read this month before you need to register a free LeadDev.com account.
Estimated reading time: 3 minutes
The provider of several popular IDEs wants to give users a choice over AI models and agents, but recent pricing changes have frustrated users.
Although JetBrains, the maker of a popular set of integrated development environments (IDEs), offers its own AI coding agent called Junie, it has recently introduced another, Anthropic’s Claude Agent, via an AI chat interface. Claude Agent has full access to IDE capabilities through the JetBrains Model Context Protocol (MCP) server.
JetBrains has also joined Google and Zed Industries in adopting the fledgling Agent Client Protocol (ACP), a standard for how AI agents interact with code editors and IDEs. JetBrains says it wants to add more agents beyond Claude Code and Junie, so users will be able to easily switch between multiple agents with a single AI subscription.
JetBrains also offers Koog for building AI agents with Kotlin, and two proprietary Large Language Models (LLMs): Grazie for technical writers, and Mellum for code completion.
The question is whether it makes sense for the toolmaker to attempt to compete with the likes of Anthropic and OpenAI by training its own models, or to focus on developer tools and use third party integrations for AI.
JetBrains is experimenting with both, and is seeking to harvest data from its customers including code snippets, prompt text, AI responses, edit history, and terminal usage, to train its proprietary LLMs. “That sounds like a lot, and it is, but that’s where the real value for improvements comes from,” according to the official post.
Your inbox, upgraded.
Receive weekly engineering insights to level up your leadership approach.
What JetBrains uses
Junie runs on ChatGPT-5 internally by default (with Claude 4.5, 4, or 3.7 available as alternatives). This model is separate from the Claude Agent feature, which users access through the AI chat interface.
Junie proved popular at launch but recent pricing changes have frustrated users. Beyond the basic free tier, which still requires you to provide a credit card, a subscription for ‘AI Ultimate’ is $30 per month and includes 35 AI Credits per 30 days, which the firm advertises as suitable for daily usage.
Super users in revolt
StocksToTrade’s hands-on CTO, Jamil Ben Alluch, disagrees. He bought an annual AI Ultimate subscription and, prior to the changes, would burn through his AI credits in about three weeks. “Now it’s about ten times more expensive; I burn through them in about four hours on GPT5, and Anthropic is even worse,” he told LeadDev.
“I feel it is unfair to do that to anyone who paid for a year’s subscription. We’re stuck with a credit limitation we weren’t planning for.”
He is far from alone. Amongst reviews of Junie in the Jetbrains marketplace, Marcelo Gomes Jr, writes: “I feel like it’s usually better than Github Copilot Pro, but the way the quota works makes this mostly unusable.”
Daniel März writes: “When you contact support … they simply ignore the fact that you purchased a product for a year and that the performance was reduced to about one-fifth or one-sixth in the middle of the contract period. This borders on fraud and will certainly be the last time I purchase an AI product from JetBrains.” There are plenty of similarly negative comments amongst the reviews.
Alluch has switched to using Claude Code, he told LeadDev. “It’s more expensive but I’m only paying for what I consume, and I don’t have to constantly top up the credits I’m using,” he says.
In response to the backlash, JetBrains published a FAQ, unapologetically stating: “There’s no scam here. […] Discounts vary, and yes, some companies still burn VC money to attract users. […] We’re a real business; we can’t play that game.”
Having built its reputation on developer-friendly tools, JetBrains faces the challenge of proving it can navigate the AI era without alienating the community that made it successful.
It remains to be seen whether developers will accept the trade-off of sharing their code and usage patterns for potentially better AI assistance, or whether they’ll migrate to competitors offering more transparent pricing and unified products.
