# ModelsLab MCP server

> Generate images, video, speech, music, and run LLM chat using over 10,000 AI models from any MCP client.

- Repository: https://github.com/ModelsLab/modelslab-mcp
- Page: https://hostedmcp.org/servers/modelslab/modelslab-mcp
- Transport: remote
- Stars: 0
- License: MIT
- Latest version: 1.0.0

## Install

### Claude Code

```sh
claude mcp add --transport http modelslab-mcp https://modelslab.com/mcp/v7
```

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

```json
{
  "mcpServers": {
    "modelslab-mcp": {
      "url": "https://modelslab.com/mcp/v7"
    }
  }
}
```

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

```json
{
  "servers": {
    "modelslab-mcp": {
      "type": "http",
      "url": "https://modelslab.com/mcp/v7"
    }
  }
}
```

### VS Code CLI

```sh
code --add-mcp '{"name":"modelslab-mcp","type":"http","url":"https://modelslab.com/mcp/v7"}'
```

### Codex

```sh
codex mcp add modelslab-mcp --url https://modelslab.com/mcp/v7
```

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

```toml
[mcp_servers.modelslab-mcp]
url = "https://modelslab.com/mcp/v7"
```

### Gemini CLI

```sh
gemini mcp add -s user --transport http modelslab-mcp https://modelslab.com/mcp/v7
```

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

```json
{
  "mcpServers": {
    "modelslab-mcp": {
      "url": "https://modelslab.com/mcp/v7",
      "type": "http"
    }
  }
}
```

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `MODELSLAB_API_KEY` | no | no |  |

## FAQ

### How do I authenticate with ModelsLab?

You can use OAuth 2.1 with PKCE or provide an API key via an Authorization Bearer header.

### How does the generation process work?

Most generation tasks are asynchronous; the tool returns a processing status and an ID, which you then use with the fetch-generation tool to check for completion.

### What kind of models are available?

The platform provides access to over 10,000 models including Flux, Qwen Image, Kling, Veo, Wan, and various open-source LLMs.

### Does this require local installation?

No, ModelsLab runs a hosted, remote MCP server that you simply point your client to via a URL.
