# genome MCP server

> Provides fully local, air-gapped semantic memory for AI agents with deterministic storage and point-in-time fact retrieval.

- Repository: https://github.com/NORTHTEKDevs/genome
- Page: https://hostedmcp.org/servers/northtekdevs/genome
- Transport: stdio
- Stars: 12
- License: Apache-2.0
- Latest version: 1.3.2

## Install

### Claude Code

```sh
claude mcp add genome -- uvx genome-memory
```

### Cursor (`~/.cursor/mcp.json`)

```json
{
  "mcpServers": {
    "genome": {
      "command": "uvx",
      "args": [
        "genome-memory"
      ]
    }
  }
}
```

### VS Code (`.vscode/mcp.json`)

```json
{
  "servers": {
    "genome": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "genome-memory"
      ]
    }
  }
}
```

### VS Code CLI

```sh
code --add-mcp '{"name":"genome","type":"stdio","command":"uvx","args":["genome-memory"]}'
```

### Claude Desktop (`claude_desktop_config.json`)

```json
{
  "mcpServers": {
    "genome": {
      "command": "uvx",
      "args": [
        "genome-memory"
      ]
    }
  }
}
```

### Codex

```sh
codex mcp add genome -- uvx genome-memory
```

### Codex config.toml (`~/.codex/config.toml`)

```toml
[mcp_servers.genome]
command = "uvx"
args = ["genome-memory"]
```

### Gemini CLI

```sh
gemini mcp add -s user genome uvx -- genome-memory
```

### Gemini settings.json (`~/.gemini/settings.json`)

```json
{
  "mcpServers": {
    "genome": {
      "command": "uvx",
      "args": [
        "genome-memory"
      ]
    }
  }
}
```

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `GENOME_ALLOW_NO_AUTH` | no | no |  |
| `GENOME_API_KEY` | no | no |  |

## FAQ

### Does genome require an LLM or API key to store memories?

No, the write path uses a local embedder, requiring zero LLM calls and zero network calls to store messages.

### How do I use genome with Claude Desktop or Cursor?

You can add it as an MCP server by installing the package and adding the command to your configuration.