JetBrains Unveils AI Agent IDE from Abandoned Fleet

JetBrains Launches Air, Agentic AI IDE Built on Abandoned Fleet

Key Facts

What: JetBrains previewed Air, an agentic AI development environment that runs multiple AI agents concurrently to handle coding tasks.
When: Public preview launched March 10, 2026; only macOS version currently available, with Windows and Linux support promised later.
Base: Air is built on the Fleet IDE codebase, which JetBrains officially discontinued as a standalone product with downloads ending December 22, 2025.
Supported Agents: OpenAI Codex, Anthropic Claude Agent, Google Gemini CLI, and JetBrains Junie; uses the vendor-neutral Agent Client Protocol (ACP).
Pricing: Junie CLI starts at $10 per month for individuals and $60 per month for enterprise licenses.

Lead paragraph

JetBrains has launched Air, a new agentic AI-powered development environment that lets developers delegate tasks to multiple AI agents running in parallel. The tool, now in public preview, is constructed on the codebase of Fleet, the company's previously abandoned next-generation IDE that was discontinued in December 2025. Air represents JetBrains' attempt to adapt to the rapidly evolving AI-assisted coding landscape while maintaining relevance amid growing competition from model providers building their own developer tools.

Background on Fleet's Demise

Fleet was originally conceived as a lightweight, modern developer environment intended to eventually complement or replace JetBrains' long-standing IntelliJ IDEA platform, which traces its roots to a Java IDE first released in 2001. Despite years in preview, Fleet failed to gain sufficient traction as a standalone product.

In December 2025, JetBrains announced it would cease supporting Fleet as an independent offering. Downloads became unavailable after December 22, 2025, although existing installations would continue to function with gradually degrading features, particularly those dependent on JetBrains servers or the AI Assistant.

The company acknowledged that maintaining two overlapping IDE families aimed at largely the same audience had become unsustainable. "Fleet did not succeed as a standalone product. We could neither replace IntelliJ IDEA with Fleet nor narrow it into a clear, differentiated niche," JetBrains stated in its official Fleet blog post.

Many technical components from Fleet have been repurposed across other JetBrains IDEs, but the core vision of Fleet as a distinct product was retired in favor of focusing on agentic development.

Air: Designed for Concurrent AI Agents

According to Nik Tkachev, JetBrains Head of Product, Air embodies what the company sees as a new wave of development tooling centered around AI agents. "Agents are changing how software is made," Tkachev said in the announcement.

The core concept in Air is the "task" — a user-described objective that one or more AI agents can execute. Tasks can run directly in a local workspace, a Git worktree, Docker container, or, in a future release, a cloud container. A built-in code editor allows developers to review, switch between tasks, and approve or modify agent output.

Air currently supports several leading AI coding agents and models, including OpenAI Codex, Anthropic's Claude Agent, Google Gemini CLI, and JetBrains' own Junie AI. It implements the Agent Client Protocol (ACP), a vendor-neutral standard for agent-editor communication that JetBrains co-sponsors with the Zed team. This approach is intended to ensure future compatibility with any compliant AI agents.

The public preview currently offers only a macOS version. JetBrains has committed to releasing Windows and Linux builds in the coming months.

Junie CLI Becomes Standalone

Alongside Air, JetBrains released Junie CLI, transforming its Junie AI offering from an IDE extension into a fully standalone command-line tool. Junie CLI supports macOS, Linux, and Windows.

Users can purchase JetBrains AI tokens or bring their own API keys for models from OpenAI, Anthropic, Google, and Grok (xAI). Pricing begins at $10 per month for individual developers and rises to $60 per month for enterprise licenses.

Some developers have requested support for locally hosted models such as those running via Ollama or Qwen. JetBrains responded that while there is no current estimated time of arrival, support for local models remains "an active topic in our planning."

Market Context and Challenges

The rise of AI-assisted coding and now fully agentic workflows has dramatically reshaped the integrated development environment market. Traditional IDEs like IntelliJ IDEA face pressure as large language model providers, including Anthropic and OpenAI, introduce their own specialized coding tools and interfaces.

"New concepts are emerging faster than anyone can validate them," Tkachev noted, highlighting the challenge of keeping pace with rapid innovation in the space.

JetBrains maintains a loyal user base for its family of IDEs built on the IntelliJ platform. However, the company now faces the complex task of balancing the needs of existing customers — many of whom have expressed little interest in migrating to an entirely new environment — while attempting to establish Air as a forward-looking, AI-centric development platform.

The decision to repurpose Fleet's codebase for Air has drawn mixed reactions. Some developers view it as an efficient use of existing technology, while others see it as evidence that JetBrains is playing catch-up in the agentic AI space rather than leading innovation.

Impact on Developers and the Industry

For developers already invested in the JetBrains ecosystem, Air offers a potential path to more automated workflows without completely abandoning familiar tooling philosophies. The ability to run multiple agents concurrently on separate tasks could significantly accelerate certain development activities, particularly for larger projects or when handling parallel workstreams like feature implementation, testing, and documentation.

The inclusion of the Agent Client Protocol suggests JetBrains is prioritizing openness and interoperability, which may appeal to teams wanting flexibility across different AI providers rather than being locked into a single vendor's ecosystem.

However, the product's current limitations — macOS-only preview, lack of local model support, and its foundation on a previously discontinued project — may give some developers pause. IntelliJ loyalists in particular are watching closely to see whether Air represents a genuine evolution or a diversion of resources away from the tools they depend on daily.

What's Next

JetBrains has not provided a detailed roadmap or timeline for full release of Air across all platforms. The company has indicated that cloud container support is planned for a future update.

Support for local models via Ollama and similar tools remains under consideration but without a committed schedule. The evolution of the Agent Client Protocol will likely play a significant role in determining how extensible Air becomes.

The broader industry continues to experiment with various approaches to agentic development. Whether JetBrains can successfully position Air as a compelling alternative to both traditional IDEs and the growing number of AI-native coding environments from model providers will be a key story to watch in 2026.

Impact Section

Air arrives at a pivotal moment for the developer tools industry. As AI agents take on more responsibility for writing, reviewing, and refactoring code, the role of the IDE itself is being redefined. Traditional feature-rich environments optimized for human programmers may need to evolve into orchestration platforms capable of managing fleets of specialized AI agents.

For JetBrains, success with Air could help secure its position in an increasingly fragmented market. Failure to gain meaningful adoption, however, might reinforce perceptions that the company is more comfortable iterating on its established IntelliJ-based products than pioneering the next generation of AI-first development environments.

Enterprise teams, in particular, will be evaluating whether Air's task-based approach and multi-agent capabilities deliver measurable productivity gains that justify adopting yet another development tool alongside their existing JetBrains IDEs.

JetBrains launches AI agent IDE built on the corpse of abandoned Fleet

Sources

Original Source

Related Topics

Comments