# OpenMemory RAG MCP Server

MCP server for importing documents into OpenMemory RAG knowledge base. Supports file uploads, URL imports, and text content ingestion.

## Features

- 📄 **Import Files**: PDF, DOCX, TXT, MD, HTML
- 🌐 **Import URLs**: Webpages, articles, documentation
- 📝 **Import Text**: Raw text content
- 🔍 **Search**: Query the knowledge base

## Installation

```bash
npm install
npm run build
```

## Configuration

Set environment variables:

```bash
export OPENMEMORY_URL="http://localhost:8080"
export OPENMEMORY_USER_ID="rag_user"
```

## Usage

### With Claude Desktop

Add to `~/.config/claude/claude_desktop_config.json`:

```json
{
  "mcpServers": {
    "openmemory-rag": {
      "command": "node",
      "args": ["/path/to/openmemory-rag-mcp/dist/index.js"],
      "env": {
        "OPENMEMORY_URL": "http://localhost:8080",
        "OPENMEMORY_USER_ID": "my_knowledge_base"
      }
    }
  }
}
```

### With Cursor

Add to `~/.cursor/mcp.json`:

```json
{
  "mcpServers": {
    "openmemory-rag": {
      "command": "node",
      "args": ["/path/to/openmemory-rag-mcp/dist/index.js"],
      "env": {
        "OPENMEMORY_URL": "http://localhost:8080"
      }
    }
  }
}
```

### With Windsurf

Add to `~/.windsurf/mcp.json`:

```json
{
  "mcpServers": {
    "openmemory-rag": {
      "command": "node",
      "args": ["/path/to/openmemory-rag-mcp/dist/index.js"]
    }
  }
}
```

## Available Tools

### 1. import_file

Import a local file into the knowledge base.

**Example:**
```
"Import the file /home/user/document.pdf into my knowledge base"
```

**Parameters:**
- `file_path` (required): Absolute path to the file
- `user_id` (optional): User ID for the knowledge base
- `tags` (optional): Array of tags

### 2. import_url

Import content from a URL.

**Example:**
```
"Import this article: https://example.com/article"
```

**Parameters:**
- `url` (required): URL to import
- `user_id` (optional): User ID
- `tags` (optional): Array of tags

### 3. import_text

Import raw text content.

**Example:**
```
"Save this to my knowledge base: [your text content]"
```

**Parameters:**
- `content` (required): Text content
- `title` (optional): Title for the content
- `user_id` (optional): User ID
- `tags` (optional): Array of tags

### 4. search_knowledge

Search the knowledge base.

**Example:**
```
"Search my knowledge base for information about Python"
```

**Parameters:**
- `query` (required): Search query
- `user_id` (optional): User ID to search within
- `limit` (optional): Maximum results (default: 5)

## Prerequisites

- OpenMemory backend running on `http://localhost:8080`
- Node.js 18+

## Quick Start

1. **Start OpenMemory backend:**
```bash
cd /path/to/OpenMemory/packages/openmemory-js
npm run dev
```

2. **Build MCP server:**
```bash
cd openmemory-rag-mcp
npm install
npm run build
```

3. **Configure AI tool** (see Usage section above)

4. **Restart AI tool** (Claude/Cursor/Windsurf)

5. **Test:**
```
"Import the file /home/user/notes.pdf"
"Search for Python programming"
```

## Supported File Types

| Type | Extensions |
|------|------------|
| Documents | `.pdf`, `.docx`, `.txt`, `.md`, `.html` |
| Web | Any URL |
| Text | Raw text content |

## Example Workflow

```
User: "Import my research paper at /home/user/research.pdf"
AI: ✅ File imported successfully!
    Memory ID: abc-123
    Tokens: 5000
    Strategy: root-child
    Sections: 3

User: "What does my research paper say about machine learning?"
AI: 🔍 Found 2 results for "machine learning":
    1. [Score: 0.892] Machine learning is a subset of artificial intelligence...
    2. [Score: 0.845] The paper discusses various ML algorithms including...
```

## Troubleshooting

### MCP tools not showing up
- Ensure OpenMemory backend is running: `curl http://localhost:8080/health`
- Check MCP config file path is correct
- Restart AI tool after configuration

### Import fails
- Verify file path is absolute
- Check file permissions
- Ensure OpenMemory backend is accessible

### Search returns no results
- Verify `user_id` matches the one used during import
- Check if content was actually imported
- Try broader search terms

## Development

```bash
# Watch mode
npm run watch

# Build
npm run build

# Run directly
npm start
```

## License

MIT