# Claude Code Setup Guide for MemoVault
This guide explains how to configure MemoVault as an MCP server for Claude Code.
## Prerequisites
1. Install MemoVault:
```bash
pip install memovault
```
2. Set up your environment:
```bash
# Create .env file or export variables
export MEMOVAULT_LLM_BACKEND=openai
export MEMOVAULT_OPENAI_API_KEY=sk-your-key-here
export MEMOVAULT_OPENAI_MODEL=gpt-4o-mini
```
## Configuration
### Option 1: Global Claude Code Settings
Add to your Claude Code settings file (`~/.config/claude-code/settings.json` or equivalent):
```json
{
"mcpServers": {
"memovault": {
"command": "memovault-mcp",
"env": {
"MEMOVAULT_LLM_BACKEND": "openai",
"MEMOVAULT_OPENAI_API_KEY": "sk-your-key-here",
"MEMOVAULT_OPENAI_MODEL": "gpt-4o-mini",
"MEMOVAULT_EMBEDDER_BACKEND": "openai",
"MEMOVAULT_EMBEDDER_OPENAI_MODEL": "text-embedding-3-small"
}
}
}
}
```
### Option 2: Project-specific Configuration
Add to your project's `.claude/settings.json`:
```json
{
"mcpServers": {
"memovault": {
"command": "python",
"args": ["-m", "memovault.api.mcp"],
"env": {
"MEMOVAULT_DATA_DIR": "./project_memories",
"MEMOVAULT_LLM_BACKEND": "openai",
"MEMOVAULT_OPENAI_API_KEY": "sk-your-key-here"
}
}
}
}
```
### Option 3: Using Local Ollama
For fully local operation:
```json
{
"mcpServers": {
"memovault": {
"command": "memovault-mcp",
"env": {
"MEMOVAULT_LLM_BACKEND": "ollama",
"MEMOVAULT_OLLAMA_MODEL": "llama3.1:latest",
"MEMOVAULT_EMBEDDER_BACKEND": "ollama",
"MEMOVAULT_EMBEDDER_OLLAMA_MODEL": "nomic-embed-text:latest"
}
}
}
}
```
## Usage in Claude Code
Once configured, you can use natural language to interact with your memories:
### Storing Memories
- "Remember that I prefer Python for backend development"
- "Store this: My project deadline is March 15th"
- "Add to memory: I like dark mode in all applications"
### Recalling Memories
- "What do you remember about my preferences?"
- "Search for anything about my projects"
- "What's my favorite programming language?"
### Memory-Enhanced Chat
- "Given what you know about me, what language should I use for this backend?"
- "Based on my preferences, suggest some VS Code themes"
## Available MCP Tools
| Tool | Description |
|------|-------------|
| `add_memory` | Store new information |
| `search_memories` | Find relevant memories |
| `chat_with_memory` | Memory-enhanced chat |
| `get_memory` | Retrieve specific memory by ID |
| `delete_memory` | Remove a memory |
| `list_memories` | Show recent memories |
| `clear_memories` | Clear all memories |
| `memory_status` | Check system status |
## Data Persistence
By default, memories are stored in `./memovault_data`. Configure with:
```bash
export MEMOVAULT_DATA_DIR=/path/to/your/data
```
## Troubleshooting
### MCP Server Not Starting
1. Check that MemoVault is installed: `pip show memovault`
2. Test the server manually: `memovault-mcp`
3. Check environment variables are set correctly
### Memory Search Not Working
1. Ensure embedder backend is configured
2. For OpenAI: verify API key is valid
3. For Ollama: ensure the embedding model is pulled
### Slow Performance
1. Consider using a smaller embedding model
2. For Ollama: ensure GPU acceleration is enabled
3. Limit `top_k` for searches
## Example Session
```
User: Remember that I prefer TypeScript for frontend projects
