Junie has been amazing for me, completely replaced my payments for Claude Code and Cursor. And it was free (until today). It's the least aggressive agent i've used, no complete re-writes or even close, And is able to achieve about 95% of what I ask of it
The only downside - which might be fixed in the newest release - is that it completely forgets context between messages - even in the same message window. But that feels like both cost cutting and easy to fix
My biggest issues with Claude Code and Cursor for what its worth:
Claude Code: Price, plus when it doesnt get things right, within a few messages it ALWAYS just creates a new file entry point with some demo console.logs that do nothing but show messages, and claims to have succeeded in what I asked
Cursor: Will break some functionality in my web application while fixing or creating others, about 80% of the time
You can add a .noai file to the root of your project to disable AI support.
https://www.jetbrains.com/help/idea/disable-ai-assistant.htm...
Errr, that 404s now. I hope the functionality still works.
Here's the cached version:
https://web.archive.org/web/20250329143832/https://www.jetbr...
> Restrict usage of AI Assistant for a project Create an empty file named .noai in the root directory of the project.
> When this file is present, all AI Assistant features are fully disabled for the project. Even if this project is opened in another IDE, the AI Assistant features will not be available.
I used JetBrains AI for about a year, it was pretty good to basically help me to scaffold things, it felt like instructing a Junior developer, which isn't bad, saved me time for side projects.
Some observations from the pricingpage:
https://www.jetbrains.com/ai-ides/buy/?section=personal&bill...
* They say free for all IDEs except the community version of PyCharm and IntelliJ.
* Looks like if you want to use your own LLM you need to be an enterprise user? None of the lower tiers allow for it, I find this really, really dumb, if I'm paying for compute, why can't I also run my own LLM? Am I misunderstanding this?
* ReSharper and Android Studio don't fall under the credit system? I really would like to know what that means.
Summary:
- JetBrains AI tools (AI Assistant and Junie coding agent) are now available under a single subscription starting with version 2025.1.
- There are three tiers: AI Free, AI Pro, and AI Ultimate.
- The free tier offers unlimited code completion, local AI models, and credit-based access to cloud AI assistance and Junie.
- AI Assistant supports Claude 3.7 Sonnet and Google Gemini 2.5 Pro.
- Junie is powered by Anthropic's Claude and OpenAI models.
It's not clear whether AI Free will be available in Community Edition IDEs or not.
Update: From the AI Plans & Pricing page, there's a tooltip that says: "The free tier is not available in the Community Editions of PyCharm and IntelliJ IDEA."
"JetBrains Junie is now publicly available to all our IDE users."
Only for IntelliJ, WebStorm and PyCharm, so far.
I've been paying for the basic AI Assistant for months now, and am glad to see a bit more clarity around this.
I got confused by the Go in the title; "But their Go IDE is called GoLand?" They should not have capitalized it.
Anyway, has anyone compared Junie with competing products?
The free tier now supports connecting to local AI models running on LM Studio or Ollama, but it still doesn't actually function without an internet connection.
If you block access to the internet or to their AI API servers [1], it refuses to start a new chat invocation. If you block access halfway through a conversation, the conversation continues just fine, so there's no technical barrier to them actually running offline, they just don't allow it.
Their settings page also says that they can't even guarantee that they implemented the offline toggle properly, a flag that should be the easiest thing in the world to enforce:
>Prevents most remote calls, prioritizing local models. Despite these safeguards, rare instances of cloud usage may still occur.
So you can't even block access to the very servers that they say their faulty offline toggle would leak data to.
[1] https://www.jetbrains.com/help/ai-assistant/disable-ai-assis...
> The AI Free tier gives you unlimited code completion and access to local AI models
Looking forward to giving this a try.
Work provides me with tooling and requires that I stick to approved AI tools, and my hobby-coding alone is just not important or regular enough to justify a paid subscription.
It's been a little annoying that I can have ollama running locally, enable ollama and configure it in my IDE, but still (seemingly?) not be able to make use of it without activating a paid AI Assistant license.
It makes perfect sense that cloud models would require payment, and that JetBrains would make some margin on that.
But I'm already paying for an IDE whose headline features have recently been so AI-focused, and if I'm also providing the compute, then I should really be able to use those features.
> On top of all that, the All Products Pack and dotUltimate subscriptions will now come with AI Pro included.
Well, colour me surprised. I've used JetBrains as an an example of a pretty decent company in the past (e.g. the way they remind you your subscription is up for renewal a couple of months in advance, so you have all the time in the world to unsub if you like), but I wasn't expecting them to just add this to the existing subscriptions.
Oh hey, they've worked some more on the features: https://www.jetbrains.com/ai-assistant/
And it's available with the all products pack: https://www.jetbrains.com/ai-ides/buy/?section=personal&bill...
I think I'll be more than happy to try it out then cause I have that pack and compare it against the Github Copilot plugin, or the likes of Continue.dev (which was pretty good in VS Code, but kind of buggy in JetBrains IDEs).
How does JetBrains AI compare to Cursor?
Separate note...
I don't know when "AI Pro" became bundled with the "all products pack" but I don't think it was there last year. I've been paying for it separately for a bit now, but a couple days ago I noticed 2 licenses showing up in the 'AI Assistant' tab. Was able to just cancel the monthly one, and use the one from 'all products pack'. May look to upgrade again when the Junie stuff becomes available in the other IDEs.
I'd like to know more about what is powering Junie under the hood.
> According to SWEBench Verified, a curated benchmark of 500 developer tasks, Junie can solve 53.6% of tasks on a single run
That's nice I guess, but why isn't this an entry on the actual https://www.swebench.com/#verified website? (Also: 53% isn't that impressive these days, Claude Sonnet can reach 63%)
I have been playing with it for a couple of hours. I have the All Product Pack which includes the AI Pro tier for free. You can hook up local models easily as well using Ollama or LM Studio. This seems better than both Continue.dev and Copilot. I will probably be cancelling my copilot subscription before the current year is up.
I set 'codebase off' but it keeps adding random files as attachment!?! This (?) also makes requests VERY slow!
Early days! Code changes break by including meta information, like:
The provided snippet is a modification of the original code, with some changes …
Here's the complete code file after applying these changes: ```python
IMO I much prefer to have my dev tools be completely FOSS so I'm not building my career on skills that are tied to a proprietary sw provider.
Does JetBrains have an equivalent of Next Edit Suggestions (VSCode) / Edit Prediction (Zed) / "Tab, tab, tab" (Cursor)?
Any MCP? … Yes and it works great!
(except that it's currently pretty slow probably because servers are overloaded)
Except for Junie. MCP not available for/in Junie?
I just downloaded IntelliJ community edition yesterday to try Kotlin. I wonder if this applies to it or only to premium version.
I currently have the $100/yr license for IntelliJ ai integration. I assume nothing changes? (Very happy with it)
Dumb question? How come in PyCharm when I go to update, it only offers me v2024.3.5 and not v2025.1?
Does IntelliJ provide coding models that run offline? Or third party provider should be used?
Great --now it’s going to use even more memory.
Since the ukraine war, jet brains products have gone way down hill. Their best engineers were in ukraine. The IDEs have become buggy and slow, sad to see this once very polished software lose relevance
I'd be a lot more excited about this if I weren't paying for a crappy VSCode clone.
Happy I never jumped on the JetBrains bandwagon...
I was curious about Z̶i̶g̶ Zed, until they too started adding AI garbage. Sad.
Luckily, I can be fairly confident that my trusty neovim will never add AI garbage unless I specifically go out of my way to install plug-ins.
EDIT: Zed, not Zig
It's pretty surprising that they're willing to charge a flat rate rather than by token, but great news for users. It's inevitable that you get annoyed at AI when it consumes tokens and generates a bad answer, or starts reading files that aren't fully relevant. The flat rate takes away that bad taste. The business modelling behind it must be quite intense, I hope this doesn't blow up in JetBrains' face if Junie's usage patterns change over time.
JetBrains are in a great position to do this though, perhaps the best position. Whereas a tool like Claude Code or Aider can give the LLM grep and little else [1], Junie can give the LLM a kind of textual API to the IDE's own static analysis database. If Claude/GPT want to understand what a function is doing and how it's used, it could issue a tool call that brings up nicely rendered API docs and nothing else, it could issue a tool call to navigate into it and read just the function body and so on. And they can use the IDE to check whether the code complies with the language rules more or less instantly without needing to do a full compile to pick up on hallucinated APIs or syntax errors.
So much potential with this kind of integration, all that stuff is barely the start.
[1] Aider attempts to build a "repo map" using a PageRank over symbols extracted by tree-sitter, but it never worked well for me.