timem-mcp 0.1.0

TiMEM MCP Server - Model Context Protocol server for TiMEM Engine

TiMEM MCP Server

Model Context Protocol server for TiMEM Engine, providing memory management tools for AI applications.

Installation

pip install timem-mcp

Usage

With Claude Desktop

Add to your Claude Desktop configuration (claude_desktop_config.json):

{
  "mcpServers": {
    "TiMEM-MCP": {
      "command": "uvx",
      "args": ["timem-mcp"],
      "env": {
        "TiMEM_API_KEY": "<your-api-key>",
        "TiMEM_API_HOST": "https://api.timem.cloud",  // optional
        "TiMEM_USER_ID": "<your-user-id>"
      }
    }
  }
}

Direct Execution

# Using uvx (recommended)
uvx timem-mcp

# Or using Python module
python -m timem_mcp

# Or directly if installed
timem-mcp

Configuration

The server reads configuration from environment variables:

Variable	Required	Default	Description
TiMEM_API_KEY	Yes	-	Your TiMEM Engine API key
TiMEM_USER_ID	Yes	-	User identifier for memory operations
TiMEM_API_HOST	No	https://api.timem.cloud	TiMEM Engine API endpoint

Note: Both TiMEM_* and TIMEM_* prefixes are supported. TiMEM_* takes priority.

Available Tools

create_memory

Create memories from conversation history.

Parameters:

• messages (required): List of message objects with role and content
• session_id (required): Session identifier
• expert_id (optional): Expert ID, default "default"
• domain (optional): Business domain, default "general"

Example:

await create_memory(
    messages=[
        {"role": "user", "content": "My name is John"},
        {"role": "assistant", "content": "Hello John!"}
    ],
    session_id="sess_123",
    expert_id="default",
    domain="general"
)

search_memories

Search and retrieve stored memories.

Parameters:

• query (optional): Search keywords
• layer (optional): Memory layer (L1-L5)
• domain (optional): Business domain
• limit (optional): Result count, default 10

Example:

await search_memories(query="John", layer="L1", limit=5)

ready

Health check endpoint to verify the server is running.

Development

# Install in development mode
pip install -e .

# Run tests
pytest

# Run with MCP Inspector
npx @modelcontextprotocol/inspector uvx --from <your-local-path> timem-mcp -e TiMEM_API_HOST=http://localhost:8000 -e TiMEM_API_KEY=<your-api-key> -e TiMEM_USER_ID=<your-user-id>

License

MIT License - see LICENSE file for details.