Google's Antigravity Renders Jupyter Notebooks Without Any Setup
Antigravity's IDE now renders .ipynb files directly in the file viewer, showing code, Markdown, tables, images, and outputs without external tools.
- Antigravity now renders
.ipynbfiles natively, including code, Markdown, tables, images, and outputs - Feature is view-only: no built-in cell execution, extensions still needed to run code
- Ships alongside remote cloud agents and Gemini Live voice mode, both rolling out gradually
- Removes the need to install Jupyter/Python marketplace extensions just to inspect notebooks
- Terminal launch via
jupyter notebook --no-browserstill works for execution workflows - Available now in the public preview build of Google Antigravity, free to use
Google’s agent-first IDE, Antigravity, can now open Jupyter .ipynb files in its native file viewer. The viewer renders saved code cells, Markdown, tables, images, and cell outputs without requiring the Jupyter and Python marketplace extensions.
Before this update, developers had to install both extensions, select a Python kernel, or open a notebook server to inspect a notebook inside Antigravity. Native rendering removes that setup for read-only review while preserving the existing execution options.
The same release redesigns the Home and Automations pages and adds download management and browsing history to the built-in browser. Google is also gradually rolling out Gemini Live voice mode and remote agents, which can continue conversations and tasks in the cloud.
Rendering stops short of execution
The native viewer displays content already stored in a notebook file. It cannot run cells, recompute outputs, or manage a kernel, so saved charts and metrics may be missing or outdated if the notebook was not executed before its last save.
| Task | Native viewer | Additional setup |
|---|---|---|
| Read code and Markdown | Supported | None |
| View saved tables, images, and outputs | Supported | None |
| Run cells | Unsupported | Jupyter extension and kernel, or a notebook server |
| Let an agent edit and execute notebooks | Unsupported | A compatible MCP integration |
Developers who need execution can still start Jupyter from Antigravity’s integrated terminal:
jupyter notebook --no-browser --port=8888Antigravity detects the running server and can open it in a browser tab or embedded view. The server manages kernels and cell execution, while the native file viewer provides quick access to the notebook’s saved state.
Faster inspection for data projects
Antigravity centers its workflow on agents that plan, write, and run code across a repository. Jupyter notebooks are common in machine learning, data analysis, and research projects, where code often appears beside charts, explanations, and experimental results. Rendering those files inside the IDE gives developers a readable view of that context.
The viewer supports several common review tasks:
- Inspect training metrics, plots, tables, and other saved outputs.
- Read notebook code and Markdown without navigating raw notebook JSON.
- Open example notebooks referenced by project documentation.
- Review notebook contents without switching to JupyterLab or another editor.
MCP bridges can add agent execution
Community projects such as antigravity-nb expose notebook operations through the Model Context Protocol, or MCP. These integrations can let an Antigravity agent read, edit, and execute notebook cells through an external server.
An MCP bridge runs with access to local files and notebook kernels, so teams should review its source, permissions, and command execution behavior before connecting it to a workspace. Google’s native viewer does not depend on an MCP server for rendering.
Platforms and rollout
The notebook viewer is available in the current Antigravity preview for macOS, Windows, and Linux. Antigravity remains free during public preview, with usage limits associated with the developer’s Google AI account.
Google lists the notebook viewer and related release details in its official changelog. Gemini Live voice mode and cloud-based remote agents are rolling out gradually, so their availability may vary by account.
Developers who only need to inspect notebooks can now do so without installing Jupyter tooling. Workflows that edit or execute cells still require a Python kernel, a running notebook server, or a third-party MCP integration.
Comments
