AI coding assistant that brings Claude Code integration, agent mode, inline chat, and auto-complete to JupyterLab notebooks. Supports multiple LLM providers and includes built-in tools for notebook editing and execution.
An AI coding assistant and extensible AI framework for JupyterLab that supports Claude Code integration, multiple LLM providers, and advanced features like agent mode, inline chat, and auto-complete.
This skill helps users install and configure Notebook Intelligence (NBI), a powerful JupyterLab extension that brings AI-powered coding assistance directly into notebooks. It supports Claude Code integration with all its features (tools, skills, MCP servers, custom commands), as well as GitHub Copilot and any other LLM provider including local models from Ollama.
Check that JupyterLab 4.0.0 or higher is installed:
```bash
jupyter lab --version
```
If not installed or version is too old:
```bash
pip install --upgrade jupyterlab
```
Install the NBI extension:
```bash
pip install notebook-intelligence
```
After installation, restart JupyterLab to activate the extension:
```bash
jupyter lab
```
1. Open JupyterLab and click the gear icon in the NBI Chat UI, or go to **Settings** → **Notebook Intelligence Settings**
2. Toggle **Enable Claude mode**
3. Configure the following options:
- **Chat model**: Select Claude model for Agent Chat UI and inline chat (e.g., claude-sonnet-4-5)
- **Auto-complete model**: Select Claude model for auto-complete suggestions
- **Chat Agent setting sources**: Choose user settings, project settings, or both for Claude Code integration
- **Chat Agent tools**: Enable Claude Code tools and/or Jupyter UI tools
- **API key**: Enter your Claude API key from https://console.anthropic.com
- **Base URL**: Enter Claude base URL (leave default unless using custom endpoint)
To use GitHub Copilot, Ollama, or other LLM providers:
1. Open **Settings** → **Notebook Intelligence Settings**
2. Select your preferred provider from the provider list
3. Configure model options and API credentials
4. For GitHub Copilot: Check **Remember my GitHub Copilot access token** to avoid re-login
To integrate MCP servers, edit the configuration file:
```bash
mkdir -p ~/.jupyter/nbi
nano ~/.jupyter/nbi/mcp.json
```
Add server configuration (example):
```json
{
"mcpServers": {
"filesystem": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-filesystem", "/path/to/directory"]
}
}
}
```
Restart JupyterLab for MCP changes to take effect.
Create custom rules for coding standards and best practices:
```bash
mkdir -p ~/.jupyter/nbi/rules
cat > ~/.jupyter/nbi/rules/01-coding-standards.md << 'EOF'
---
priority: 10
---
EOF
echo '{"rules_enabled": true}' > ~/.jupyter/nbi/config.json
```
For mode-specific rules (agent, ask, inline-chat):
```bash
mkdir -p ~/.jupyter/nbi/rules/modes/agent
nano ~/.jupyter/nbi/rules/modes/agent/01-testing.md
```
1. Open NBI Chat UI in JupyterLab
2. Type a request like "Create a notebook that analyzes the iris dataset"
3. The AI agent will create, edit, and execute cells interactively
1. In a notebook cell, click the sparkle icon or press **Ctrl/Cmd + G**
2. Type your code generation request
3. Press **Ctrl/Cmd + Enter** to accept the suggestion
1. Start typing in a code cell or Python file
2. Suggestions appear automatically
3. Press **Tab** to accept
1. Configure MCP servers in `~/.jupyter/nbi/mcp.json`
2. In Agent mode, reference MCP tools with `@mcp` participant
3. The agent can discover and use tools from configured MCP servers
To disable specific built-in tools, edit JupyterLab config:
```python
c.NotebookIntelligence.disabled_tools = ["nbi-notebook-execute", "nbi-python-file-edit"]
```
Available tool IDs:
1. **Extension not appearing**: Restart JupyterLab after installation
2. **Claude API errors**: Verify API key in Settings and check network connectivity
3. **MCP servers not working**: Ensure model supports tool calling; restart JupyterLab after config changes
4. **Rules not applying**: Check `rules_enabled: true` in config.json; verify file paths and YAML frontmatter syntax
Leave a review
No reviews yet. Be the first to review this skill!
# Download SKILL.md from killerskills.ai/api/skills/notebook-intelligence-for-jupyterlab/raw