TaoDock AI Image & Video MCP Server

Use TaoDock image and video models from Claude Desktop, Cursor, VS Code, Gemini CLI, and other local MCP clients. One TaoDock API key discovers the models authorized for your account, generates images, starts asynchronous video jobs, checks their status, and downloads completed media.

The server uses local stdio transport. It does not expose a remote HTTP or SSE endpoint. Image and video generation consume TaoDock quota and generated files are written to your computer.

中文简介：在 Claude Desktop、Cursor、VS Code、Gemini CLI 等 MCP 客户端中调用 TaoDock 图片与视频模型。详细隐私边界见 PRIVACY.md。

mcp-name: io.github.taodock-api/mcp-server

Requirements

Node.js 20.11 or later
A TaoDock account and API key
An MCP client that can launch a local stdio server

Create an account at taodock.com/sign-up, review your wallet, then create a key at taodock.com/keys with the models you need.

Quick start

Use the same server definition in any MCP client that accepts mcpServers JSON:

{
  "mcpServers": {
    "taodock": {
      "command": "npx",
      "args": ["-y", "@taodock-ai/mcp-server@0.2.0"],
      "env": {
        "TAODOCK_API_KEY": "<your TaoDock API key>",
        "TAODOCK_OUTPUT_DIR": "/absolute/path/to/TaoDock"
      }
    }
  }
}

Pinning the package version makes client startup reproducible. Remove @0.2.0 only if you intentionally want every launch to use the latest release.

Claude Desktop

Add the configuration above to:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\\Claude\\claude_desktop_config.json

Restart Claude Desktop after saving. You can also install the signed release artifact from GitHub Releases: taodock-mcp-0.2.0.mcpb.

Cursor

Open Settings → Tools & Integrations → MCP Tools → New MCP Server, then add the JSON definition. Cursor also supports project configuration in .cursor/mcp.json and user configuration in ~/.cursor/mcp.json.

VS Code

Create .vscode/mcp.json in a workspace or use the user-level MCP configuration:

{
  "servers": {
    "taodock": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@taodock-ai/mcp-server@0.2.0"],
      "env": {
        "TAODOCK_API_KEY": "${input:taodock-api-key}",
        "TAODOCK_OUTPUT_DIR": "/absolute/path/to/TaoDock"
      }
    }
  },
  "inputs": [
    {
      "id": "taodock-api-key",
      "type": "promptString",
      "description": "TaoDock API key",
      "password": true
    }
  ]
}

Gemini CLI

Add a local MCP server with Gemini CLI:

gemini mcp add --scope user \
  -e TAODOCK_API_KEY=<your-key> \
  -e TAODOCK_OUTPUT_DIR=/absolute/path/to/TaoDock \
  taodock npx -y @taodock-ai/mcp-server@0.2.0

Configuration

Variable	Required	Description
`TAODOCK_API_KEY`	Yes	TaoDock API key from taodock.com/keys
`TAODOCK_BASE_URL`	No	Trusted HTTPS-compatible endpoint; defaults to `https://taodock.com/v1`. The API key is sent to this endpoint.
`TAODOCK_OUTPUT_DIR`	No	Directory for generated files. Use an absolute path; otherwise the default is `./taodock-output` relative to the MCP process.
`TAODOCK_LOCALE`	No	Error language: `en` or `zh`; defaults to `en`

Never configure an untrusted base URL. The server rejects plain HTTP and URLs containing embedded credentials.

Tools

Tool	Purpose	Side effects
`list_models`	List image, video, or text models available to the configured key	Reads the authorized model list and public catalog
`generate_image`	Generate an image and save the result	Consumes quota and writes local files
`create_video`	Start an asynchronous video generation task	Consumes quota and creates a remote task
`get_video_status`	Read a video task's current status	Read-only API request
`download_video`	Wait for and save a completed video	Writes a local file

Image-specific fields belong in parameters; video-specific fields belong in request. The selected model and prompt cannot be overridden by these objects. Consult the TaoDock model guide for supported fields.

Example prompts

List the image models available to my TaoDock key.
Generate a 1024×1024 product hero image with a clean white background.
Create a video task with an available video model and return the task ID immediately.
Check video task <task-id>, then download it when completed.

Review usage and costs in TaoDock usage logs.

Run from source

git clone https://github.com/taodock-api/taodock-mcp.git
cd taodock-mcp
npm ci
npm run validate
TAODOCK_API_KEY=<your-key> npm start

For local client development, replace the npx command with an absolute path to Node and dist/index.js. Run npm run mcpb to validate and build the Claude Desktop extension.

Security and privacy

API keys are sent only as Bearer credentials to the configured HTTPS API endpoint.
Generated image URLs must use HTTPS, redirects are rejected, responses must be images, and downloads are limited to 25 MiB.
Generated media remains in TAODOCK_OUTPUT_DIR until you delete it.
The MCP host may retain prompts, tool results, file paths, or logs according to its settings.
See PRIVACY.md and the TaoDock privacy policy.

Report vulnerabilities privately through the repository's GitHub security advisory page. Do not include API keys in public issues.

Troubleshooting

401: replace the missing, expired, disabled, or invalid key in API Keys.
model_not_found: verify the key's group and model permissions, then check the model catalog.
403 / insufficient_user_quota: check the key quota and wallet.
Empty image or video list: confirm that the model is enabled for the key and supports the required endpoint.
Server exits immediately: use Node.js 20.11 or later and inspect the MCP client's stderr logs.
Cannot find generated files: set TAODOCK_OUTPUT_DIR to an absolute path.

TaoDock integrations

Documentation: https://docs.taodock.com · Support: https://github.com/taodock-api/taodock-mcp/issues

mcp-server

Documentation