The Pulse
JetBrains Adds Air for Multi-Agent Workflows to PyCharm 2026.3
JetBrains says PyCharm 2026.3 ships with Air, a layer for running and reviewing multi-agent coding workflows inside the IDE. The release also adds subproject-specific Python consoles and terminals, FastAPI dependency graphs, deeper Pydantic
AI.info Team · Written with AI · fact-checked ·
JetBrains says PyCharm 2026.3 ships with Air, a new layer for agentic development that lets developers run, track and verify multi-agent workflows inside the IDE. JetBrains says code intelligence helps agents make fewer wrong turns, while review tools let developers examine every change before it ships. The release also brings updates to Python workspaces, language support and framework tooling.
JetBrains says Air works with existing agents and subscriptions
JetBrains says Air works with Claude Agent, Codex, Gemini, GitHub Copilot, Junie, OpenCode and other agents compatible with ACP. It detects agents already installed on a developer’s machine, which can be connected in a click, and lets users add agents from a registry. Developers can work with the coding agents of their choice.
Existing provider subscriptions, API keys and custom base URLs work with Air, and JetBrains says no JetBrains AI subscription is required. Users can run Claude Agent through a Claude subscription in the terminal or a full-screen IDE tab, and use other agent subscriptions in chat or the terminal. They can switch subscriptions at any moment, even mid-session.
Air also includes free AI with the Junie Lite configuration, which requires a JetBrains Account but no subscription. For developers moving from AI Assistant, Air automatically migrates IDE skills, MCP configuration, agents installed through ACP, authentication and chat history. Chat mode history is not migrated.
JetBrains describes a central workspace for reviewing agent changes
Air lets developers review an agent’s changes with the surrounding code and project tools in view. Developers can send feedback directly to an agent, anchor a note to a line in a diff or comment from the code, so the agent gets concrete context about the change wanted. The interface supports work in either chat or the terminal.
Developers using AI Assistant will receive a notification offering an Air trial. The release post says moving to Air hides AI Assistant, and points users to instructions for returning to it.
Each PyCharm subproject gets its own console
In PyCharm 2026.3, each subproject gets a Python console and terminal running in its own environment. That lets developers run code against the dependencies used by that subproject. In uv workspaces, the console uses the workspace environment and adds the subproject root to sys.path.
Users can open a subproject’s console by right-clicking the subproject, or launch a terminal while a file from that subproject is open in the editor. They can also create a subproject or convert an existing folder from the Project tool window. PyCharm initializes the subproject with uv, Hatch or Poetry.
For uv workspaces, PyCharm automatically adds a newly initialized subproject as a workspace member in the root pyproject.toml. The redesigned Python interpreter widget detects environment tools installed on the machine, including uv, Poetry, Hatch and conda. It lists environments managed by those tools for the project, including environments created from the command line.
Initial Python 3.15 lazy import support and FastAPI and Pydantic tools
PyCharm adds initial support for the lazy import syntax introduced in Python 3.15. Lazy imports defer loading a module until its first use, and PyCharm parses and resolves them like regular imports. The IDE also adds interactive dependency graphs for FastAPI endpoints in the Pro edition.
FastAPI graphs show chains of Depends() and Security() calls across files, mark dependencies as yield or return, and flag circular dependencies in the editor. A dependency view can show which endpoints rely on it, directly or through other dependencies. Users can open a graph from the gutter beside an endpoint or view its usages from the endpoint count above a dependency.
PyCharm’s deeper Pydantic support reads field constraints, validators and configuration from model declarations. It checks values passed to a model against constraints such as min_length, max_length, gt and lt, and warns when a constraint is broken. Field names in @field_validator('price') resolve to their definitions, while completion suggests model configuration options such as validate_by_name and validate_by_alias.