MCP Digital Human Server

A Model Context Protocol (MCP) server for the AceDataCloud Digital Human API.

Features

• Create lip-synced digital human videos from a source face video or image
• Clone voices from short reference audio samples
• Poll, batch-retrieve, or delete Digital Human tasks

Connect locally

Use the local stdio package and an AceDataCloud API token. The public hosted HTTPS endpoint could not be verified, so this README does not offer it as a connection option. The local MCP process still calls AceDataCloud's API; it does not run the underlying service offline.

Get an API credential through AceDataCloud Platform, keep it out of committed files, and review current pricing before a billable call.

Local stdio

Install the package and give the local process an API token:

python -m pip install mcp-digitalhuman
export ACEDATACLOUD_API_TOKEN='YOUR_API_TOKEN'
mcp-digitalhuman

For Claude Desktop local MCP, merge this entry into the file opened by its developer settings (~/Library/Application Support/Claude/claude_desktop_config.json on macOS). uvx requires uv on PATH:

{
  "mcpServers": {
    "digitalhuman": {
      "command": "uvx",
      "args": ["mcp-digitalhuman"],
      "env": {"ACEDATACLOUD_API_TOKEN": "YOUR_API_TOKEN"}
    }
  }
}

Keep this user-level file private. Self-hosted HTTP uses mcp-digitalhuman --transport http --port 8000; expose it only with suitable network and TLS controls. Local execution still calls the AceDataCloud API.

Check the connection

Confirm that the MCP client loads its tools. For an existing task ID, digitalhuman_get_task makes an authenticated query; loading the tool alone does not prove downstream API access. A real service request can be billed. If it returns 401, check the API token; if it returns 403 or an account/balance error, read that response before retrying.

Development

pip install -e ".[dev,test]"
pytest -q
ruff check .

Documentation

Documentation