# Model Ledger MCP server

> Tracks deployed models and their dependencies using an append-only log to provide a connected dependency graph for AI agents.

- Repository: https://github.com/block/model-ledger
- Page: https://hostedmcp.org/servers/block/model-ledger
- Transport: stdio
- Stars: 17
- License: Apache-2.0
- Latest version: 0.7.13

## Install

### Claude Code

```sh
claude mcp add model-ledger -- uvx model-ledger
```

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

```json
{
  "mcpServers": {
    "model-ledger": {
      "command": "uvx",
      "args": [
        "model-ledger"
      ]
    }
  }
}
```

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

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

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

```json
{
  "mcpServers": {
    "model-ledger": {
      "command": "uvx",
      "args": [
        "model-ledger"
      ]
    }
  }
}
```

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

```toml
[mcp_servers.model-ledger]
command = "uvx"
args = ["model-ledger"]
```

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

```json
{
  "mcpServers": {
    "model-ledger": {
      "command": "uvx",
      "args": [
        "model-ledger"
      ]
    }
  }
}
```

## FAQ

### What makes Model Ledger different from MLflow or SageMaker?

Unlike registries tied to a single platform, Model Ledger spans all platforms as one connected graph.

### How does the dependency graph work?

The graph builds automatically when an output port name matches an input port name, requiring no manual wiring.

### Can I use this with Claude?

Yes, the MCP server allows you to point Claude or any MCP agent at your model inventory.