JetBrains Air: Constructing a System of Products for Agentic Software program Growth
AI
JetBrains
New Products
News
Releases
AI can produce code. Organizations nonetheless have to supply software program. Agentic growth is altering how software program will get made, but it surely hasn’t modified what it prices to be flawed.
Six months in the past, we started publicly experimenting with agentic growth environments. Around the identical time, we launched JetBrains Central as an open management and execution system for agent-driven growth. We subsequently started rolling out JetBrains Central CLI, shared context, cloud brokers, automations, governance, and AI price controls for groups and organizations.
Today, we’re bringing this work collectively as JetBrains Air: an open, coherent system of merchandise for builders, groups, and organizations, inside and past JetBrains IDEs. It is multi-surface and multi-service. Each product solves a definite drawback, however the merchandise work higher collectively.
JetBrains Air marks a big growth in what JetBrains is constructing for. For 26 years, we’ve targeted totally on the person developer workbench. Now, we’re constructing for the broader system by way of which agentic work is initiated, executed, coordinated, reviewed, and ruled.
The IDE stays essential to JetBrains’ future. The period during which the entire software program growth system might be contained in a single window is ending. As a part of our continued capital allocation, we at the moment are bringing the foundational agentic expertise into JetBrains IDEs, giving skilled builders an setting the place they’ll work successfully with brokers whereas understanding, altering, and verifying the ensuing code. JetBrains Air connects the broader system growing round it.
That system relies on a core perception that the way forward for agentic growth shall be multi-vendor. No single mannequin, agent, or service shall be proper for each developer, crew, or activity.
From one product to an open system of merchandise
This strategic shift has a sensible consequence: JetBrains Air can’t be simply one other agent or growth setting. It should join merchandise for particular person work, crew coordination, organizational management, context, and course of automation – and stay open to the instruments and brokers builders select, together with these JetBrains doesn’t construct.
JetBrains Air contains merchandise which might be accessible right this moment alongside others that shall be launched because the system develops:
- Air in JetBrains IDEs – an entire agentic growth expertise for guiding and orchestrating brokers and verifying their work inside JetBrains IDEs.
- Air Teams – a brand new solution to coordinate and automate software-delivery workflows involving builders and autonomous brokers.
- Air Governance (previously JetBrains Central) – organizational coverage, visibility, auditability, price administration, and accountability for AI-assisted and agent-driven growth.
Junie is JetBrains’ coding agent for skilled software program growth. It shall be supported throughout all Air surfaces.
Air in JetBrains IDEs provides builders the setting to direct brokers and confirm their output utilizing JetBrains’ code intelligence. Air Teams turns particular person agent exercise into coordinated crew workflows. Air Governance makes that exercise seen, governable, and accountable throughout the group.
But an open system can not cease at JetBrains’ personal merchandise. The Agent Client Protocol (ACP) standardizes the connection between an IDE and an agent’s full harness, together with its planning, logic, instruments, mannequin routing, and observability. Through the ACP Registry, builders can uncover and run a rising vary of suitable brokers whereas persevering with to work inside JetBrains IDEs.
Air Governance is designed to increase visibility and price governance throughout suppliers and the completely different instruments by way of which agentic work takes place. This means builders can select the agent, mannequin, or service suited to the duty with out forcing the group to surrender context, visibility, or management.
Together, the Air merchandise enable work to maneuver between builders, brokers, instruments, and environments with out dropping the context and controls surrounding it.
Individual adoption has moved sooner than organizational utilities
Since March, our merchandise have progressed considerably, however so has our understanding of what agentic growth requires.
Developers have been adopting brokers sooner than organizations can construct the utilities round them. Agent capabilities have superior, and completely different fashions and brokers have confirmed helpful for various duties. However, the context, coordination, governance, and price administration surrounding them haven’t saved tempo.
For many builders, brokers are already delivering sensible worth. At the organizational degree, the economics are a lot more durable to show. The prices floor elsewhere – in evaluation, rework, safety, utilities, and spend.
Which brokers can entry firm code? Where can knowledge go? Which output requires human evaluation? What occurred whereas an agent was working remotely? Who accredited the ensuing change, and the way was it verified?
Fragmentation at this degree isn’t simply irritating. It makes software program growth more durable to know, measure, and govern at precisely the purpose when extra of the work is being delegated.
The bottleneck is shifting with the work
Code that’s clearly flawed will get caught shortly. That a part of the system nonetheless works. The more durable drawback is code that’s virtually proper: believable, able to passing a superficial verify, however quietly carrying a nasty assumption or architectural inconsistency that gained’t floor till it’s costly.
As brokers tackle extra of the execution, the bottleneck shifts from producing change to understanding, verifying, and proudly owning it. Code turns into cheaper to generate however costlier to confirm. Agent exercise turns into simpler to begin however more durable to coordinate, audit, and clarify.
And whereas the work might be delegated, accountability can not. An agent is not going to get the decision at 3:00 am when one thing breaks. The duty for what ships nonetheless belongs to the individuals and organizations that ship it.
This is why management turns into more durable, not simpler, as AI improves. A extra succesful mannequin might produce higher output. It doesn’t set up organizational coverage, protect provenance, present price visibility, or resolve who accepts duty for the ensuing change.
The future is multi-vendor
Multi-vendor assist is a foundational design precept of JetBrains Air, shaping how the system is being constructed from the outset.
We don’t imagine this market will consolidate any time quickly. Models range in what they’re good at, and rankings change each few months. Teams inside the identical firm already make completely different decisions, and they’re usually proper to take action. Standardizing on one AI vendor right this moment means making a multi-year dedication in a market that gained’t look the identical subsequent quarter.
Keeping the choices open is the affordable factor to do. The drawback is what openness presently prices. Every new mannequin, agent, or service a corporation provides takes away somewhat extra visibility into its personal growth work. Context doesn’t carry over between instruments. Spend can’t be attributed. Policies need to be rebuilt for every service.
Organizations mustn’t have to decide on between utilizing one of the best accessible instruments and understanding what is occurring inside their very own engineering. That trade-off exists as a result of nothing within the present stack was constructed to sit down above a number of distributors directly.
This is the work JetBrains has taken on. We construct our personal agent, and we intend to make it glorious. But JetBrains Air doesn’t require prospects to make use of ours, and our technique doesn’t rely upon which mannequin supplier leads the rankings this quarter. We haven’t any purpose to make the ecosystem smaller than it’s.
What we will provide as a substitute is one place to run, see, govern, and account for agentic growth throughout each mannequin, agent, and repair – for the developer, the crew, and the group.
Supporting a number of fashions and brokers is the ground, not the ceiling. The half that issues is what sits above them: shared context, one set of insurance policies, a single price view, and a file of what occurred, no matter which vendor produced the change.
Why JetBrains?
Multi-vendor selection solves solely a part of the issue. Agents additionally want dependable software program intelligence.
JetBrains brings 26 years of engineering intelligence to the issue, serving to builders perceive the construction and conduct of complicated software program, not merely generate extra of it. That deterministic code intelligence offers a basis for making agentic work extra dependable, environment friendly, and comprehensible throughout completely different fashions and brokers. We are seeing promising outcomes from giving AI brokers entry to deterministic code intelligence.
This is an financial benefit in addition to a technical one. Agents spend money and time rediscovering info the codebase already accommodates. An agent that may retrieve that data is cheaper and extra correct than one which has to reconstruct it. Because intelligence doesn’t belong to 1 mannequin, the profit can lengthen throughout supported brokers and providers.
We are additionally going by way of the identical transition because the organizations we construct for, adopting brokers internally, redesigning workflows, and studying the place particular person productiveness positive factors translate into higher software program supply and the place they merely transfer work elsewhere.
What comes subsequent
JetBrains Air will develop by way of a rolling sequence of releases. We shall be specific about what prospects can use now, what’s getting into preview, and what stays a part of our longer-term path.
Over time, JetBrains Air will lengthen additional into cellular and distant experiences, permitting individuals to provoke, monitor, evaluation, and proceed agentic work because it strikes between environments. The purpose is to not reproduce the IDE on each floor. We are making the best context and controls accessible wherever selections should be made.
We may even deliver JetBrains’ intelligence into extra agentic workflows. This contains richer context drawn from code, structure, repositories, runtime conduct, and organizational data, in addition to higher methods to route work between builders, fashions, brokers, and providers.
More work shall be triggered by repository occasions, schedules, and supply processes reasonably than by a developer opening an editor and issuing a immediate. JetBrains Air will present the intelligence, oversight, and human management these workflows require throughout surfaces and providers.
We is not going to identify future merchandise earlier than their scope and availability are able to be confirmed. With every launch, we are going to clarify what works, the way it connects, and what’s nonetheless in progress.
Where JetBrains Air goes
The firms that reach adopting AI is not going to essentially be those who generate probably the most code or deploy probably the most brokers. They shall be these that may increase experimentation with out dropping high quality, context, price self-discipline, or human understanding.
JetBrains Air is our dedication to constructing for that actuality. It expands JetBrains from the developer workbench right into a system of merchandise connecting builders, brokers, groups, and organizations.
The purpose isn’t extra code. It is software program that builders, groups, and organizations can perceive, confirm, and stand behind.
