JetBrains announced the availability of PyCharm 2026.2.1, with a set of updates focused on AI-assisted software development and improved handling of Python environments and interactive notebooks. Agents such as Claude Code and Codex can now create, edit, and run Jupyter notebooks through PyCharm’s notebook interface using a live runtime kernel.
This approach allows variables, models, and data to remain available between cells, rather than running separate scripts that do not share the same state. According to JetBrains, this enables notebook and machine-learning tasks to be executed more reliably while using fewer tokens. To get started, users can open an AI chat and ask it to perform a task inside the notebook.
Directing Agents to the Correct Python Environment
The release introduces a new skill called Agent environment coordinator, which provides the AI agent with information about the interpreter and tool configured for the project. Supported tools include uv, Poetry, pip within a virtual environment, and conda, so run or package-installation commands target the environment associated with the project rather than the system environment.
If no environment is configured, one can be created through PyCharm, while the agent decides how to use the available information. Users can enable this by asking the agent to run or install something in the project.
Support for marimo Notebooks and PyCharm Plugins
marimo notebooks can now be opened, edited, and run directly inside PyCharm through a new plugin developed by the marimo team. These notebooks support interactive cells and interactive user-interface elements within a dedicated notebook without leaving the development environment.
Because marimo notebooks are stored as Python files, they are compatible with Git and can be run as scripts and integrated into existing Python projects. The plugin is available for separate installation.
As part of maintaining bundled plugins, JetBrains has removed from the default package or stopped bundling low-usage plugins, including Data Wrangler, Hugging Face, and Google Colab support. Compatible versions can still be installed from the JetBrains Marketplace, but these plugins are no longer bundled or actively supported by the PyCharm team. The goal of this step is to provide a more focused set of bundled plugins and help keep PyCharm responsive.
Python Package Management and Type Checking
The Python Packages tool window has been redesigned. Packages now appear in a collapsible tree alongside their dependencies, with new icons, right-aligned version displays, and direct links for installation and updating.
A dedicated floating package search window has also been added, allowing users to find and install packages quickly while showing the environment in which the installation will be used. Users can select a module and dependency group, install packages within uv or Poetry dependency groups such as dev and test for each workspace member, change versions directly or through the Change Version window, and install from version-control repositories using Custom Installation.
Package repositories can now remember their state between sessions when enabled or disabled, while inaccessible links display a clear error message. The improvements also include support for Remote Development.
For type checking, PyCharm displays clearer and more actionable messages, with richer details for type-mismatch errors, a type difference for functions and compound types, and fully displayed names and types in inspection tooltips with clickable links.
Fixes and Availability
PyCharm no longer creates virtual environments for end-of-life Python versions: Python 2.7, 3.6, and 3.7. It remains possible to create a new environment from the command line and add it manually to the development environment. In addition, inference for SQLAlchemy 2.0 Session.get(), as well as SQLModel, now treats Session.get(Entity, id) as an instance of the model rather than the class.
All these updates are available in PyCharm 2026.2.1. Users can update from within the development environment or through the Toolbox app, or download the release to try it on their own projects.