- Python 3.11+ - Required for all functionality
- uv - Modern Python package and project manager
-
Clone and install dependencies:
git clone git@github.com:memovai/memov.git cd memov # Basic version (core features only) uv sync # With VectorDB/RAG support (semantic search, validation, debugging) uv sync --extra rag
-
Set up development tools:
uv pip install pre-commit uv run pre-commit install
-
Optional: Link core memory package (for advanced development):
# Clone the core mem package cd .. git clone git@github.com:memovai/mem.git # Install as editable dependency cd memov uv pip install -e ../mem
After uv sync, you can test your local MCP server with different clients:
# Navigate to your test project directory
cd /path/to/your/test/project
# Add local dev MCP server
claude mcp add mem-mcp --scope project -- /path/to/memov/.venv/bin/mem-mcp-launcher stdio $(pwd)Create .vscode/mcp.json in your test project:
{
"servers": {
"mem-mcp": {
"type": "stdio",
"command": "/path/to/memov/.venv/bin/mem-mcp-launcher",
"args": ["stdio", "${workspaceFolder}"]
}
}
}Go to Files > Preferences > Cursor Settings > MCP, then add:
{
"mcpServers": {
"mem-mcp": {
"command": "/path/to/memov/.venv/bin/mem-mcp-launcher",
"args": ["stdio", "${workspaceFolder}"]
}
}
}You can also use uv run from the memov directory:
# From the memov project directory
uv run mem-mcp-launcher stdio /path/to/your/test/projectThe project provides two main entry points:
| Command | Purpose | Module |
|---|---|---|
mem-mcp-server |
Server management CLI | mem_mcp_server.cli.server_cli:main |
mem-mcp-launcher |
Direct MCP runtime launcher | mem_mcp_server.server.mcp_launcher:main |
Run commands without global installation:
uv run mem-mcp-server --help
uv run mem-mcp-launcher --helpDefault config directory: ~/.mem_mcp_server
~/.mem_mcp_server/
└── logs/ # Server log files