JetBrains IDEs 2026.3 bundle Air coding-agent workspace | Magica
JetBrains bundles Air into IDEs; standalone app is in maintenance
Editorial Team
••📖7 min read
JetBrains IDEs 2026.3 bundle Air for running coding agents and reviewing their changes, while the standalone Air app is in maintenance mode. Existing provider subscriptions remain usable, but billing, security controls and execution options vary by integration.
Air is bundled from JetBrains IDEs 2026.3; it was an optional Marketplace installation in 2026.2.
The standalone Air app is in maintenance mode and retains Docker support that the IDE plugin lacks.
Existing agent accounts can be used without a JetBrains AI subscription, subject to provider and interface restrictions.
Copilot's dedicated plugin retains capabilities absent from its Air integration, including a local sandbox and GitHub enterprise controls.
JetBrains, the developer-tools company behind integrated development environments (IDEs) such as PhpStorm for PHP and RubyMine for Ruby, is bundling its Air coding-agent workspace with the 2026.3 releases. Air lets developers delegate tasks to agents and review their output inside the tools they use to inspect, debug and commit code.
The move comes as JetBrains says in a support note updated November 3 that its earlier standalone Air desktop app is in maintenance mode, with no new feature updates planned. Bringing the workspace into established IDEs changes where developers supervise agents, but does not preserve every standalone capability or make competing agent integrations equivalent.
JetBrains’ release screenshot shows Air’s project and session list alongside a coding task and its changes. Source: JetBrains. Source: JetBrains.
Agent choice predates the bundled workspace
The company's 2026.3 release overview lists support for Claude Agent, Codex, Gemini, GitHub Copilot, Junie, OpenCode and other agents compatible with the Agent Client Protocol, or ACP. Air detects agents installed on the developer's machine and offers a registry for adding others.
Agent choice and subscription reuse are not new to this release. In January, JetBrains and code-editor maker Zed announced a shared ACP agent registry integrated into JetBrains IDEs and Zed. JetBrains' instructions already let developers install agents through AI Assistant, its existing plugin for chat, completion and other AI features, and use them without a JetBrains AI subscription. ACP provides a common interface between coding agents and supporting editors, reducing the need for a dedicated integration for each pairing.
Amazon’s shares and convertible notes could be worth $271.8 billion at the top of Anthropic’s proposed IPO range. Existing accounting gains, a 12% ownership cap and Anthropic’s computing commitments shape what that paper value means.
In March, JetBrains launched a public preview of standalone Air for delegating tasks to multiple agents concurrently. That announcement described an environment built around supervising agents, with the IDE handling the rest of development. The bundled plugin now brings that workspace into the IDE itself.
JetBrains' documentation says Air requires installation in 2026.2 but is bundled starting with 2026.3, carries no separate plugin license and is available in JetBrains IDEs except Android Studio. JetBrains still describes it as under active development, with rough edges and frequent changes.
JetBrains distinguishes Air's session-focused workspace from AI Assistant's companion features inside the editor. AI Assistant also supports coding agents and ACP installations; Air emphasizes parallel sessions, session diffs, reusable presets for agent, model and operating mode, and dedicated session management. Developers can inspect changes alongside surrounding code, use refactoring and debugging tools, and edit or revert results. Feedback can be attached to a line in a diff or sent from the editor.
Language-specific tools can supply another check. A report on PhpStorm's release describes inspections accessible to agents through its Model Context Protocol (MCP) server and a bundled skill that gives agents the project's PHP interpreter. The interpreter configuration addresses a concrete failure mode: falling back to a system PHP installation with the wrong version or missing extensions. The report also describes Laravel inspections that can catch nonexistent configuration keys and methods, including through MCP.
Air’s provider settings offer JetBrains account access and third-party API-key configuration; authorization options vary by agent. Source: JetBrains. Source: JetBrains.
Existing subscriptions do not mean identical access
Air is a workspace, rather than an AI service supplying its own models. JetBrains says users can bring provider subscriptions, API keys and custom base URLs, without buying a JetBrains AI subscription. It also includes Junie Lite, a free configuration of JetBrains' own coding agent requiring a JetBrains Account.
That makes the IDE a common place to manage agents even when the underlying AI purchase remains with another provider. It also gives JetBrains' own agent an entry point alongside third-party choices. The distinction matters commercially: bundling the interface does not consolidate the user's AI bills. The support FAQ says the authorization source selected for a session determines who bills its usage.
Provider rules constrain the promised portability. A Claude Pro or Max subscription works through locally installed Claude Code in Air's terminal interface. It cannot authorize Claude Agent in the structured chat interface; JetBrains says that requires JetBrains AI or Anthropic API-key authentication with API billing. The FAQ attributes the restriction to Anthropic's terms for the SDK integration.
Data handling follows the chosen provider, too. JetBrains says code sent using a third-party subscription goes directly to that provider rather than through JetBrains, and the applicable terms and privacy policies depend on the agent or provider selected. Air does not automatically enable AI features and requires no login until an agent is authorized.
Availability also varies by provider. JetBrains says the plugin has no geographic distribution restrictions, but JetBrains AI cannot authorize sessions in China; other vendors set their own territorial limits.
Copilot shows the limits of a common interface
GitHub Copilot, GitHub's AI coding assistant, illustrates both the convenience and the compromises. Microsoft's integration announcement says developers can choose Copilot from Air's agent picker, authenticate with an existing subscription and delegate tasks without installing a separate Copilot plugin. Both routes support agentic coding, model selection, parallel sessions and IDE tools. Copilot can also supply the IDE's native Next Edit Suggestions starting with JetBrains IDE version 2026.2.4.
Microsoft nevertheless recommends the dedicated Copilot plugin for deeper integration and enterprise capabilities. Its comparison lists inline agentic chat, Copilot code review, commit-message generation, advanced agent customization and debugging, GitHub enterprise managed settings and policy diagnostics, and a local sandbox as capabilities absent from Copilot in Air. A shared agent interface therefore does not make the two experiences interchangeable.
Other integrations have limits too. JetBrains' FAQ says model selection works for Cursor's agent, but reasoning-effort, context-size and fast-mode controls are unavailable. It also lists a known bug hiding Claude and GPT models for the Antigravity agent, with a fix in progress.
Execution environments also require care. Microsoft says Copilot's Air integration extends from Local Run to Cloud Run starting November 3, with Cloud Run part of JetBrains Teams. Separately, JetBrains' FAQ says Air sessions can run locally, in temporary worktrees or in the cloud, but do not yet offer native agent execution in Docker, Dev Containers or a sandbox. Registering a containerized agent through custom ACP configuration is a workaround, with IDE tools unavailable inside that container.
This is a difference from standalone Air, whose March preview offered Docker isolation. JetBrains' November 3 support note says the desktop app retains that support despite its maintenance status. RubyMine's separate addition of native access to WSL and Dev Container projects does not establish that Air agents can execute natively in those containers.
Organizations retain controls at the JetBrains layer. Administrators can restrict agents, providers, models and MCP servers, apply policies to user groups and enforce approved-agent lists. Mandatory policies on managed machines can prevent users bypassing organizational restrictions with personal subscriptions. The documentation also distinguishes disabling AI through IDE licenses or IDE Services from revoking a Console organization workspace: the latter can leave other authorization methods available, depending on policy.
A reviewer adds a note to a changed line in Air’s diff viewer, giving the agent feedback on its code. Source: JetBrains. Source: JetBrains.
Migration and verification remain the next tests
JetBrains' migration instructions are not fully consistent. Its November 5 RubyMine announcement says installing Air uninstalls AI Assistant. The general release overview says switching hides it. The documentation says the plugins can coexist, while the support FAQ says accepting the Air notification disables AI Assistant and explains how to enable it again without removing Air.
That difference affects users who rely on AI Assistant's chat, completion and AI Actions. The release materials say Air migrates IDE skills, MCP configuration, ACP-installed agents, authentication and chat history, but excludes Chat mode history. Users should check the migration behavior and retained history in their particular IDE before treating the move as seamless. In 2026.3, the FAQ says the bundled Air plugin can be disabled but cannot be uninstalled through the ordinary plugin controls.
The immediate choice is which workflow preserves the capabilities a team needs: Air inside the IDE, the supported AI Assistant plugin, or a dedicated provider plugin such as Copilot's. Existing standalone Air users also have to weigh its retained Docker support against its maintenance status. The cited release materials offer no controlled measurement of coding speed, defect rates or total task cost. Resolving the broader productivity question requires comparing completed tasks with the same agent and model, including review effort, defects and provider charges—not simply counting parallel sessions.
Nvidia chief Jensen Huang argued in May 2025 that US chip restrictions were strengthening Chinese rivals. Filings show a substantial commercial stake, but lost sales and market share alone cannot establish whether the controls failed their security purpose.
Reported demand has Anthropic’s bankers considering a price above its filed $70–$78 range. The combined financing could exceed $100 billion, but existing cash, uneven operating results and long-term computing commitments complicate the valuation test.
The British AI chip designer has secured private and UK government backing before its planned 2027 commercial launch, putting a multibillion-dollar valuation on its promise of faster, cheaper inference.
Zhang Yiming reportedly told staff that ByteDance needs years to challenge leading US AI labs, even as the TikTok parent prepares a personal agent for release within weeks. Its consumer reach and video strengths have yet to translate into frontier language models.
Helsing has unveiled a military intelligence platform promising European control over data, models and operations. Germany's evaluation of roughly 30 European systems makes procurement and demonstrated performance the next tests.
Public safety leaders warn that synthetic voices and background sounds make hoax emergency calls harder to assess. A voluntary survey exposes preparedness gaps, but does not establish a national AI-swatting rate or measure growth over time.
Fortinet's November security disclosures include AI-assisted findings, but some affected software branches still await fixes. Earlier testing put token costs at about $540 per true positive, excluding the broader work of reviewing findings and securing customer deployments.
Tamil Nadu told the Madras High Court that accounts containing about 2.22 million rupees had been frozen after its library-procurement order, while a criminal case was still being prepared. The dispute tests safeguards ordered in 2024; AI involvement remains unconfirmed.
Unions withdrew objections to Google’s proposed $10 million purchase of Spirit Airlines data after securing employee-record exclusions and independent privacy review. The agreement follows earlier passenger-data protections; court approval and verification remain ahead.
Baird cut Nvidia to Neutral and its price target to $200 from $500. Its case combines slower forecast AI spending growth with China-linked inventory exposure and uncertainty over founder and CEO Jensen Huang’s tenure.
ByteDance is reportedly accelerating plans to combine its Spell project with Doubao. Existing travel bookings reveal a transaction-fee business, while the phone assistant's history shows how access to other apps can constrain execution.
The Office for Students wants more AI-free assessment and better AI skills teaching, backed by existing quality rules. Its research raises doubts about unsupervised coursework, but universities say assessment reform is already under way.
Finland has recognised Tesla’s provisional Dutch approval after changes to branding and speed controls. Drivers remain responsible, and the national decision does not commit Finland to backing EU-wide authorisation.
Grok Bot can connect to internal MCP servers through Tailscale without exposing them to the public internet. User-based authorization preserves network controls, but administrators must configure how narrowly the assistant can reach into their systems.
Daylight’s Goldman Sachs Alternatives-led Series B backs security agents supported by human engineers. The young company reports rapid growth, but has not disclosed the economics behind its automation promise.
Primordial Labs raised $18 million to widen deployment of Anura, its military command software. Existing contracts and demonstrations give it a foothold, but integration counts do not establish sustained use or lower operating costs.
Northwestern Polytechnical University bought two Nvidia H200 processors for 874,700 yuan and awarded five contracts for rented Nvidia computing, procurement records show. Separate US restrictions on the defense-linked institution remained in place despite January’s conditional opening for H200 exports to China.
The North Dakota Republican says allegations involving a suspected Chinese intelligence operative make Huang vulnerable to coercion. Nvidia’s board is backing its founder while independent directors investigate, with investors facing both security and leadership risks.
Scott Bessent says Washington and Beijing are establishing a channel for serious AI incidents, following their September agreement. Its usefulness depends on unresolved reporting procedures and participation by the companies that detect model failures.
