JetBrains released PyCharm 2026.3 on November 5, 2026. Like the rest of the 2026.3 IDE wave, it ships with JetBrains Air, the new layer for running AI coding agents and checking their changes from inside the IDE.
The Python-specific work is just as notable. Workspaces now give each subproject its own console and terminal, the editor understands the lazy import syntax coming in Python 3.15, and FastAPI and Pydantic projects get deeper code insight. A redesigned interpreter widget also changes how you pick and rebuild environments.
JetBrains Air comes to PyCharm
Air is a single workspace for starting agents, following multi-agent jobs and reviewing the result. PyCharm's code analysis feeds into that process, which JetBrains says helps agents avoid wrong turns before you look at the diff.
The agent list is open rather than tied to one vendor. It covers Junie, Claude Agent, Codex, Gemini, GitHub Copilot, OpenCode and any agent that supports the Agent Client Protocol (ACP). Agents already installed locally are detected automatically, and others can be added from a registry.
No JetBrains AI plan is required. Your own provider subscriptions, API keys and custom base URLs carry over, and you can switch between them during a session. Claude Agent can work in a full-screen editor tab or from the terminal, and a free Junie Lite setup needs only a JetBrains Account.
For review, the agent's edits appear next to the code around them. You can attach a note to one line of the diff or leave feedback straight from the editor, then send it back so the agent knows exactly what to change.
Users still on AI Assistant will see an offer to try Air. Moving over hides the old assistant and brings across skills, MCP settings, ACP agents, sign-in data and chat history, except history from Chat mode. JetBrains has published steps for going back.
Workspace support for multi-package Python projects
Monorepos with several Python packages get the biggest everyday change. Each subproject now gets a dedicated terminal and Python console bound to that subproject's environment, so code runs against the packages it really depends on.
- In uv workspaces, the console runs on the shared workspace environment and puts the subproject root on sys.path.
- A right-click on a subproject offers a menu entry to open a Python console for it.
- Opening the terminal while a subproject file is active starts it in that subproject's environment.
- New subprojects can be created, or existing folders converted, from the Project tool window using uv, Hatch or Poetry.
- In uv workspaces, a new subproject is registered as a member in the root pyproject.toml automatically.
Python 3.15 lazy imports, FastAPI and Pydantic
Python 3.15 adds a lazy import syntax that postpones loading a module until code first touches it. That can shorten start-up for scripts and services with heavy dependencies. PyCharm now parses and resolves these imports the same way as normal ones, so navigation and completion keep working.
FastAPI users on the Pro edition get a dependency graph for each endpoint. It follows chains of Depends() and Security() calls across files, marks each dependency as yield or return, and flags circular dependencies that would otherwise fail at start-up with a RuntimeError. A dependency can also list every endpoint that relies on it.
Pydantic support goes deeper too. PyCharm now checks values passed to a model against its Field() limits, for example gt, lt, min_length or max_length, and shows a warning on a violation. Names inside @field_validator now resolve to the field definition, and completion suggests model configuration options including validate_by_alias and validate_by_name.
A new interpreter widget and other changes
The status bar interpreter widget has been rebuilt around the tools that manage environments, including conda, Hatch, Poetry and uv. It finds the tools on your machine and groups your project's environments under each one, including environments created from the command line.
A Recreate Environment action rebuilds the current environment with another Python version, and PyCharm offers to download that version if it is missing. JetBrains also lists full SQLAlchemy ORM support, a Home window for editing single files, settings copied into new Git worktrees and SSH remote development among the wider changes.
What it means for developers
Workspace support targets a real pain point. Teams that keep several services or libraries in one uv repository often run code against the wrong environment, and per-subproject consoles and terminals should cut down on those confusing import errors.
Lazy import support lets you try the Python 3.15 feature early without losing editor help. Measure start-up time before and after on a command-line tool or a serverless function with large imports, since that is where the gain is most likely to show.
The FastAPI graph and Pydantic checks catch problems that usually appear only at run time. Air is worth trying with an agent subscription you already pay for, but keep reviewing its diffs line by line before merging anything.
How to get PyCharm 2026.3
Update from inside the IDE or download the new build from the PyCharm page on the JetBrains website. The full What's New page lists further fixes for code insight, testing and version control.
For more on this release wave, read our coverage of IntelliJ IDEA 2026.3 and GitHub Copilot in JetBrains Air.
Conclusion
PyCharm 2026.3 pairs the cross-IDE Air agent layer with solid Python work. Per-subproject environments, Python 3.15 lazy import support and smarter FastAPI and Pydantic checks make it a useful upgrade for teams with larger or multi-package codebases.
Sources
See my articles more often when you search Google.
