Kirill Skrygan retweeted
JetBrains Air is significantly better than code assistant CLIs and TUIs. ACP-based harnesses w/ ACP UIs are the IDEs of the future.
15
6
145
18,543
Very curious about the upcoming Qwen 4 27b capabilities.
If only it could be as smart as 3.8 and as fast as 3.6...
3
12
1,639
Today we have hit a new milestone for Junie Local.
It can run on your RTX devices and Windows version is stable.
100% local.
Zero costs.
Available both though Air in IDEs / Junie CLI.
More supported hardware and more powerful models coming soon!
2
6
84
3,530
We connected IntelliJ IDEA’s code intelligence to Claude Code and Codex, then tested the setup with Netflix.
The first public results are out.
Typical efficiency gains: 20–30%.
Best code-navigation case: ~65%.
These results come from two real codebases. The efficiency metric averages token savings and time savings.
Correctness stayed stable or improved.
Here’s what we actually built.
The code engine behind IntelliJ IDEA ran headlessly, giving agents access to its capabilities through a CLI backed by a daemon.
LSP alone wasn’t sufficient.
Agents needed a convenient way to invoke the tools. The backend also needed to stay running and keep the project index warm between calls.
The idea is straightforward.
An IDE has already mapped the structure of your code and its dependencies — including what’s inside compiled libraries.
An agent can use that knowledge instead of spending time and tokens reconstructing it.
But the most interesting finding was which models benefited most.
The strongest models saw the biggest gains.
Especially at higher reasoning effort.
In these experiments, more capable models were also better at taking advantage of code intelligence.
My takeaway: better models and better tools aren’t competing bets.
You don’t have to choose between investing in model intelligence and giving agents better access to your codebase.
The two can reinforce each other.
Huge thanks to @pbakker and the Netflix team for working with us on this!
39
39
294
79,612
The case study:
blog.jetbrains.com/ai/2026/1…
2
3
44
6,362
Air in my @intellijidea and the list of one-click-install agents
No, half of the list actually😀
5
2
36
3,136
Holy moly, we’ve seen a lot of interest in JetBrains Air since the announcement!
-Air in IDEs: for developers
-Air Teams: for team leads & engineering managers
-Air Governance: for CTOs & engineering directors
Here’s how to try these things:
Links below 🧵
9
10
91
27,644
Air Teams.
-Agentic automations.
-Shared cloud environments.
-JetBrains context to optimize agents in the orgs cloud.
Coming soon: mobile/slack app, autonomous personal agents.
Available now to JetBrains business customers: air.jetbrains.cloud
2
12
2,802
Air Governance.
Gives organizations visibility and control over AI agent usage: access, spending, and adoption, all in one place.
Vendor agnostic - works with MULTIPLE LLM providers (incl. BYOK)
Organization admins can get started directly at:
console.jetbrains.cloud
1
5
1,364
They say software development is dead.
Certainly for tasks like merging Diablo and StarCraft - yes.
Meanwhile, ~52% of orgs don’t even measure AI ROI :)
And those who do, just count measure token consumption, lol.
Folks, there are plenty of things to solve in SD.
6
2
89
6,697
Glad to see more and more real use cases of Air in our IDEs.
Stay tuned for more features and updates this fall😎
I tried Air in JetBrains IDEs, gave Grok, Codex, Claude & Junie different jobs, and stopped treating AI coding like a single-agent workflow.
I had them work on different parts of a Task Management System, with separate agents handling the backend, UI, testing & other parts of the implementation.
At the same time, I could switch between their sessions, see their progress, review the code, give follow-up instructions where needed & use the IDE's development tools to verify their work instead of forcing everything through one agent.
That’s the difference I underestimated. I don’t want to sit inside one coding agent writing another prompt every 5 minutes. I want to define the objective, give different agents the right tasks, let them execute in parallel, then step back in when judgment, refinement or approval is actually needed.
The old workflow was prompt → code → review → repeat.
The workflow I’m moving toward is assign → execute in parallel → verify.
Once multiple coding agents can work on different parts of a project it starts to feel less like one AI assistant and more like work you’re directing from inside your IDE.
That’s what I tested with Air in JetBrains IDEs. It lets you work with different coding agents while staying inside the JetBrains development environment to follow their work, inspect changes & decide what actually ships.
JetBrains also has its own coding agent, Junie.
For a limited time, you can use Junie with a specific model for free, with unlimited usage, to try Air in IDEs.
Try Air in JetBrains IDEs:
jb.gg/ha05m1
2
1
40
2,892
With Air in our IDEs, you can easily work with Claude, Codex, and 50+ other agents.
But we also bundled Junie Lite right into Air — so you always have a baseline agent ready to go.
And for something that’s free, it’s surprisingly good.
Check it out, share your feedback.
8
5
62
5,318
We've reimagined JetBrains IDEs to become the ultimate agentic workbench.
Air is an open hub for agents inside your IDE:
-Any agents to run (Claude, Codex & many more)
-Agents are more efficient with IDE intelligence
-Free Junie Lite agent bundled
-Cloud runs via Air Teams
⬇️
45
69
464
3,418,280
With Air Teams + Air Governance, developers in organizations can:
-Track every session's cost
-Steer cloud runs and share them
-Use any harness or model from one org-wide wallet
Mobile app, automations and more team-level features coming this fall.
2
11
2,561