MCP Memory Server

MCP Memory Server

A persistent vector memory server for Windsurf, VS Code, and other MCP-compliant editors.

Features

• Local Vectors: Uses LanceDB and all-MiniLM-L6-v2 locally. No API keys required.
• Persistence: Memories are saved to disk (./mcp_memory_data).
• Isolation: Supports multiple projects via project_id.

Installation

You need Python 3.10+ installed.

1. Setup Virtual EnvironmentIt's recommended to use a virtual environment to avoid conflicts.

   python3 -m venv .venv
   source .venv/bin/activate
   

2. Install Dependencies

   pip install -e .
   

Configuration

Add this to your mcpServers configuration (e.g., in ~/.codeium/windsurf/mcp_config.json or VS Code MCP settings).

Windsurf / VS Code Config

Replace /ABSOLUTE/PATH/TO/... with the actual path to this directory.

{
  "mcpServers": {
    "memory": {
      "command": "/ABSOLUTE/PATH/TO/mcp-memory-server/.venv/bin/python",
      "args": [
        "-m",
        "mcp_memory.server"
      ],
      "env": {
        "MCP_MEMORY_PATH": "/ABSOLUTE/PATH/TO/mcp-memory-server/mcp_memory_data"
      }
    }
  }
}

Usage

Ingestion

Use the included helper script ingest.sh to ingest context files.

# General Usage
./ingest.sh --project <PROJECT_NAME> <FILES...>

Real Example (Project "Thaama"):

/ABSOLUTE/PATH/TO/mcp-memory-server/ingest.sh --project project-thaama \
  /path/to/PROJECT_CONTEXT.md \
  /path/to/DECISIONS.md \
  /path/to/AI_RULES.md

Tools

The server exposes:

• memory_search(project_id, q)
• memory_add(project_id, id, text)

Troubleshooting

• First Run: The first time you run it, it will download the embedding model (approx 100MB). This might take a few seconds.
• Logs: Check the editor's MCP logs if connection fails.