# Claude Code Integration Guide
This guide shows how to connect the Obsidian RAG MCP server to Claude Code.
## Prerequisites
1. Clone and set up the project:
```bash
cd ~/projects/obsidian-rag-mcp
python3.12 -m venv .venv
source .venv/bin/activate
pip install -e .
```
2. Set your OpenAI API key:
```bash
export OPENAI_API_KEY=sk-your-key-here
```
3. Index your vault:
```bash
PYTHONPATH=. python -c "
from src.rag import RAGEngine
engine = RAGEngine(vault_path='./vault')
engine.index()
"
```
## Claude Code Configuration
Add this to your Claude Code MCP settings:
### Option 1: Stdio Transport (Recommended)
```json
{
"mcpServers": {
"obsidian-rag": {
"command": "/Users/lume/projects/obsidian-rag-mcp/.venv/bin/python",
"args": [
"-m", "src.mcp.server"
],
"cwd": "/Users/lume/projects/obsidian-rag-mcp",
"env": {
"OPENAI_API_KEY": "your-key-here",
"OBSIDIAN_VAULT_PATH": "/Users/lume/projects/obsidian-rag-mcp/vault",
"PYTHONPATH": "/Users/lume/projects/obsidian-rag-mcp"
}
}
}
}
```
### Option 2: Using a shell wrapper
Create `~/projects/obsidian-rag-mcp/run-server.sh`:
```bash
#!/bin/bash
cd /Users/lume/projects/obsidian-rag-mcp
source .venv/bin/activate
export OPENAI_API_KEY="your-key-here"
export OBSIDIAN_VAULT_PATH="./vault"
PYTHONPATH=. python -m src.mcp.server
```
Then in MCP config:
```json
{
"mcpServers": {
"obsidian-rag": {
"command": "/Users/lume/projects/obsidian-rag-mcp/run-server.sh"
}
}
}
```
## Available Tools
Once connected, Claude Code has access to:
| Tool | Description | Example Use |
|------|-------------|-------------|
| `search_vault` | Semantic search | "Find RCAs about database timeouts" |
| `search_by_tag` | Tag-filtered search | "Find all P1 incidents" |
| `get_note` | Read full note | "Read the billing-api runbook" |
| `get_related` | Find similar notes | "What's related to this RCA?" |
| `list_recent` | Recent changes | "Show recently modified docs" |
| `index_status` | Check index | "How many docs are indexed?" |
## Testing the Connection
In Claude Code:
```
/mcp
```
You should see `obsidian-rag` in the list of connected servers.
## Example Queries
Once connected, try:
> "Search my vault for past incidents involving CosmosDB connection issues"
> "Find RCAs similar to this kubernetes pod crash we're seeing"
> "What does our runbook say about handling billing-api memory issues?"
> "List the most recent RCA documents"
## Troubleshooting
### Server not starting
- Check Python path: `.venv/bin/python` should exist
- Verify OPENAI_API_KEY is set
- Check vault path exists
### No results
- Run `index_status` to verify the index is populated
- Re-index if needed: `engine.index(force=True)`
### Slow queries
- First query may be slow (loading models)
- Subsequent queries should be <500ms
