> ## Documentation Index
> Fetch the complete documentation index at: https://docs.pylancemcp.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Claude Desktop Setup
> Configure Pylance MCP in Claude Desktop app
## Prerequisites
* Claude Desktop app installed
* Pylance MCP Server running
* Anthropic account with API access
## Quick Start
Install via pip:
```bash theme={null}
pip install pylance-mcp-server
```
Edit Claude Desktop configuration:
**macOS**: `~/Library/Application Support/Claude/claude_desktop_config.json`
**Windows**: `%APPDATA%\Claude\claude_desktop_config.json`
**Linux**: `~/.config/Claude/claude_desktop_config.json`
```json theme={null}
{
"mcpServers": {
"pylance": {
"command": "python",
"args": ["-m", "pylance_mcp.server"],
"env": {
"PYLANCE_WORKSPACE": "/path/to/your/python/project"
}
}
}
}
```
Quit and reopen Claude Desktop app
In Claude chat, type:
```
Can you analyze my Python workspace?
```
Claude should respond with workspace information.
## Configuration Examples
### Single Workspace
```json claude_desktop_config.json theme={null}
{
"mcpServers": {
"pylance": {
"command": "python",
"args": ["-m", "pylance_mcp.server"],
"env": {
"PYLANCE_WORKSPACE": "/Users/you/projects/myapp",
"PYTHONPATH": "/Users/you/projects/myapp"
}
}
}
}
```
### Multiple Workspaces
```json claude_desktop_config.json theme={null}
{
"mcpServers": {
"pylance-backend": {
"command": "python",
"args": ["-m", "pylance_mcp.server", "--port", "3000"],
"env": {
"PYLANCE_WORKSPACE": "/Users/you/projects/backend"
}
},
"pylance-ml": {
"command": "python",
"args": ["-m", "pylance_mcp.server", "--port", "3001"],
"env": {
"PYLANCE_WORKSPACE": "/Users/you/projects/ml-models"
}
}
}
}
```
### Cloud-Hosted
```json claude_desktop_config.json theme={null}
{
"mcpServers": {
"pylance": {
"url": "https://api.pylancemcp.dev",
"headers": {
"Authorization": "Bearer YOUR_API_KEY",
"X-Workspace": "/path/to/workspace"
}
}
}
}
```
Get your API key from [pylancemcp.dev/dashboard](https://pylancemcp.dev/dashboard)
## Usage Examples
### Code Review
Ask Claude to review your code:
```
@pylance review the authentication module in src/auth.py
```
Claude will analyze:
* Type safety issues
* Potential bugs
* Security vulnerabilities
* Code style consistency
* Best practice violations
### Refactoring Assistance
Request refactoring help:
```
@pylance can you help refactor database.py to use async/await?
```
Claude will:
1. Analyze current implementation
2. Suggest refactoring approach
3. Generate refactored code
4. Explain changes made
### Documentation Generation
Generate docstrings:
```
@pylance generate docstrings for all functions in utils.py
```
Output:
```python theme={null}
def calculate_total(items: List[Item]) -> float:
"""
Calculate the total price of all items.
Args:
items: List of Item objects containing price information
Returns:
float: Sum of all item prices
Raises:
ValueError: If items list is empty
Example:
>>> items = [Item(price=10.0), Item(price=20.0)]
>>> calculate_total(items)
30.0
"""
if not items:
raise ValueError("Items list cannot be empty")
return sum(item.price for item in items)
```
### Test Generation
Generate unit tests:
```
@pylance create pytest tests for the User class in models/user.py
```
Claude generates comprehensive tests including:
* Happy path scenarios
* Edge cases
* Error handling
* Fixtures and mocks
## Advanced Features
### Context-Aware Conversations
Claude maintains context across messages:
```
You: @pylance what's the User model structure?
Claude: The User model has email, password_hash, created_at fields...
You: Add a last_login field to it
Claude: I'll add the last_login timestamp field to the User model...
```
### Multi-File Analysis
Analyze dependencies across files:
```
@pylance trace how the login function works from API endpoint to database
```
Claude maps the flow:
```
api/auth.py:login_endpoint()
→ services/auth.py:authenticate_user()
→ models/user.py:User.verify_password()
→ database.py:query_user_by_email()
```
### Intelligent Code Search
Search using natural language:
```
@pylance find all database queries that aren't using prepared statements
```
Claude locates potential SQL injection vulnerabilities.
## Conversation Templates
### Bug Investigation
```
@pylance I'm getting a TypeError on line 45 of main.py.
The error message is: "unsupported operand type(s) for +: 'int' and 'str'"
Can you help debug this?
```
### Feature Implementation
```
@pylance I need to add OAuth2 authentication to the API.
Can you:
1. Show me what files need to be modified
2. Generate the necessary code
3. Explain the security considerations
```
### Performance Optimization
```
@pylance This function is slow with large datasets:
[paste code]
Can you suggest optimizations and show benchmarks?
```
## Troubleshooting
**Error**: `Command 'python' not found`
**Fix**: Use absolute path to Python:
```json theme={null}
{
"command": "/usr/local/bin/python3",
"args": ["-m", "pylance_mcp.server"]
}
```
**Error**: `Failed to initialize workspace`
**Checklist**:
* ✅ Workspace path exists
* ✅ Path has no trailing slash
* ✅ User has read permissions
* ✅ Contains Python files (.py)
**Error**: `401 Unauthorized`
**Solutions**:
* Check API key is valid
* Verify subscription is active
* Ensure API key has correct permissions
**Symptoms**: No response to `@pylance` commands
**Debug**:
1. Check Claude Desktop logs:
* macOS: `~/Library/Logs/Claude/`
* Windows: `%APPDATA%\Claude\logs\`
2. Verify MCP server is running:
```bash theme={null}
ps aux | grep pylance_mcp
```
3. Test server manually:
```bash theme={null}
python -m pylance_mcp.server --test
```
## Performance Optimization
### Large Workspaces
For projects with >5,000 files:
```json claude_desktop_config.json theme={null}
{
"mcpServers": {
"pylance": {
"command": "python",
"args": [
"-m", "pylance_mcp.server",
"--max-workers", "4",
"--cache-dir", "/tmp/pylance-cache"
],
"env": {
"PYLANCE_WORKSPACE": "/path/to/large/project",
"PYLANCE_DIAGNOSTIC_MODE": "openFilesOnly"
}
}
}
}
```
### Memory Limits
Constrain memory usage:
```json theme={null}
{
"env": {
"PYLANCE_MAX_MEMORY": "2048", // MB
"PYLANCE_CACHE_SIZE": "500" // MB
}
}
```
## Security Best Practices
**Never** commit `claude_desktop_config.json` with API keys to version control
Use environment variables:
```json claude_desktop_config.json theme={null}
{
"mcpServers": {
"pylance": {
"url": "https://api.pylancemcp.dev",
"headers": {
"Authorization": "Bearer ${PYLANCE_API_KEY}"
}
}
}
}
```
Set in shell profile:
```bash ~/.zshrc theme={null}
export PYLANCE_API_KEY="pk_live_..."
```
## Tips & Tricks
**Tip 1**: Use `@pylance` at the start of messages for better context
**Tip 2**: Reference specific files: `@pylance analyze src/models/user.py`
**Tip 3**: Ask for alternatives: `@pylance show 3 ways to implement this function`
**Tip 4**: Request explanations: `@pylance explain this decorator pattern`
## Next Steps
Configure for VS Code
Advanced workspace configuration