[BUG] High CPU usage (~400%) when Jupyter notebook and Claude Code sidebar are both open

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Opening a Jupyter notebook in VS Code while the Claude Code sidebar is active causes CPU usage to spike to ~400%. The VS Code window process consumes excessive resources, and the developer console shows "Throttling navigation to prevent the browser from hanging" errors.

What Should Happen?

Claude Code should not cause high CPU usage when a Jupyter notebook is open. The notebook detection/connection logic should either:

1. Not run continuously in a polling loop, OR
2. Have a timeout/backoff mechanism to prevent resource exhaustion, OR
3. Only attempt connection when the user explicitly clicks "Connect"

Error Messages/Logs

Throttling navigation to prevent the browser from hanging. See https://crbug.com/1038223. Command line switch --disable-ipc-flooding-protection can be used to bypass the protection

Steps to Reproduce

1. Open VS Code with Claude Code extension installed
2. Open Claude Code sidebar
3. Open any .ipynb file (even an empty notebook)
4. Observe CPU spikes to ~400% (check via VS Code Process Explorer: Cmd+Shift+P → "Process Explorer")
5. Developer console shows: Throttling navigation to prevent the browser from hanging
6. Click the "Connect" button
7. CPU immediately returns to normal (even though the mcp__jupyter tool call fails)

Claude Model

Not sure / Multiple models

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.25

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

VS Code integrated terminal

Additional Information

Environment

• Claude Code: 2.1.25 (native install)
• Claude Code for VS Code: 2.1.25
• VS Code: 1.108.2
• Jupyter Extension: 2025.9.1
• OS: macOS 26.2 (25C56)

Key Finding

Clicking the "Connect" button immediately resolves the CPU spike, even though the tool call fails with "No such tool available: mcp__jupyter__get_cells".

This suggests Claude Code is continuously polling/attempting to detect or connect to the notebook in a loop, and the loop only exits once the Connect button is clicked (regardless of success/failure).

Additional Notes

• Opening notebook without sidebar: no CPU issue
• Opening sidebar without notebook: no CPU issue
• Both open simultaneously: CPU spikes until Connect is clicked

It is probably related to #21912 - The Connect button fails because mcp__jupyter tool is not available. However, this CPU issue affects all users who have both notebook and sidebar open, even if they don't intend to use the Jupyter integration feature.