# WARM MCP server

> Lets an AI client store, search, and recall persistent local memory with full-text and vector search plus source provenance.

- Repository: https://github.com/writerslogic/warm
- Page: https://hostedmcp.org/servers/writerslogic/warm
- Official: yes
- Transport: stdio
- Stars: 27
- License: Apache-2.0
- Latest version: 0.1.7

## Install

### Claude Code

```sh
claude mcp add warm -- npx -y writerslogic-warm-mcp
```

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

```json
{
  "mcpServers": {
    "warm": {
      "command": "npx",
      "args": [
        "-y",
        "writerslogic-warm-mcp"
      ]
    }
  }
}
```

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

```json
{
  "servers": {
    "warm": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "writerslogic-warm-mcp"
      ]
    }
  }
}
```

### VS Code CLI

```sh
code --add-mcp '{"name":"warm","type":"stdio","command":"npx","args":["-y","writerslogic-warm-mcp"]}'
```

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

```json
{
  "mcpServers": {
    "warm": {
      "command": "npx",
      "args": [
        "-y",
        "writerslogic-warm-mcp"
      ]
    }
  }
}
```

### Codex

```sh
codex mcp add warm -- npx -y writerslogic-warm-mcp
```

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

```toml
[mcp_servers.warm]
command = "npx"
args = ["-y", "writerslogic-warm-mcp"]
```

### Gemini CLI

```sh
gemini mcp add -s user warm npx -- -y writerslogic-warm-mcp
```

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

```json
{
  "mcpServers": {
    "warm": {
      "command": "npx",
      "args": [
        "-y",
        "writerslogic-warm-mcp"
      ]
    }
  }
}
```

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `WARM_API_KEY` | no | no |  |
| `WARM_URL` | no | no |  |

## FAQ

### How do I run the WARM MCP server?

Run it via npx for stdio or HTTP transport, or build the warm-mcp binary from source with the mcp feature.

### What memory operations does WARM support?

It provides remember, recall, forget and explain operations with time validity and supersession, plus source provenance.

### Does WARM work offline?

Yes, the embedded engine makes no service calls or telemetry and can run fully offline with local storage.

### Does WARM support encryption at rest?

It offers optional AES-256-GCM storage encryption with a signed audit log and COSE/SCITT provenance.

### Can I use WARM from Python or Node.js?

Yes; warm-sdk is an HTTP client for warm-server in Python and JavaScript, and writerslogic-warm embeds the Rust engine in Node.
