Model Context Protocol

Use AI services from your LLM

Connect Claude Desktop, Cursor, or any MCP-compatible client directly to Hudsonly AI. Generate speech, create avatar videos, and manage tasks without leaving your conversation.

Set Up MCP View Available Tools

What is MCP?

The Model Context Protocol is an open standard that lets AI assistants use external tools and data sources. Instead of writing code to call APIs, your LLM calls them directly.

Conversational

Ask your AI to generate speech or create videos in natural language. No code needed.

Secure

Your API token stays local. Auth, credits, and rate limits all apply the same as the REST API.

Full Feature Set

All 8 API operations are available as MCP tools, from generation to task management.

Available Tools

Your LLM gets access to these tools when connected via MCP.

generate_speech

Convert text to speech audio. Returns an audio URL immediately. 1 credit.

generate_avatar

Generate animated portrait video. Returns a task_id for polling. 10 credits.

get_task_status

Poll an async task by ID until it completes or fails.

list_tasks

List recent tasks with optional filtering by service or status.

get_credit_balance

Check your current credit balance and recent transactions.

list_services

List available AI services with descriptions and credit costs.

list_models

List available models for a service (e.g. SadTalker, LivePortrait).

get_rate_limits

Check rate limit status for each service.

Setup Guide

Connect your MCP client in two steps.

1

Get your API token

Create an API token from your dashboard or via the API. You'll need this for authentication.

2

Add the MCP server config

Add the following to your MCP client configuration. Replace YOUR_API_TOKEN with the token from step 1.

Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "hudsonly-ai": {
      "url": "https://api.ai.hudsonly.com/mcp/hudsonly",
      "headers": {
        "Authorization": "Bearer YOUR_API_TOKEN"
      }
    }
  }
}
Cursor

Add to .cursor/mcp.json in your project:

{
  "mcpServers": {
    "hudsonly-ai": {
      "url": "https://api.ai.hudsonly.com/mcp/hudsonly",
      "headers": {
        "Authorization": "Bearer YOUR_API_TOKEN"
      }
    }
  }
}
Claude Code (CLI)

Run in your terminal:

$ claude mcp add hudsonly-ai \
  --transport streamable-http \
  --url https://api.ai.hudsonly.com/mcp/hudsonly \
  --header "Authorization: Bearer YOUR_API_TOKEN"

Example Conversation

Here's what using Hudsonly AI through MCP looks like.

You

Create an audio clip that says "Thanks for watching our demo" in a natural voice, then make an avatar video with this portrait image: https://example.com/portrait.jpg

Assistant

I'll start by generating the speech audio, then use it to create the avatar video.

generate_speech({ text: "Thanks for watching our demo", voice: "af_heart" })
// Result: { audio_url: "https://storage.../audio_abc123.mp3" }
generate_avatar({ image_url: "https://example.com/portrait.jpg", audio_url: "https://storage.../audio_abc123.mp3" })
// Result: { task_id: "9f1a2b3c-...", status: "pending" }
get_task_status({ task_id: "9f1a2b3c-..." })
// Result: { status: "completed", output: { video_url: "https://storage.../video_xyz.mp4" } }

Assistant

Done! Here are your results:

  • Audio: https://storage.../audio_abc123.mp3
  • Video: https://storage.../video_xyz.mp4

The avatar video shows the portrait speaking your text with natural lip movements. Total cost: 11 credits (1 for TTS + 10 for avatar).

Compatible Clients

Works with any client that supports the MCP streamable HTTP transport.

Claude Desktop

Anthropic

Claude Code

CLI Agent

Cursor

AI Code Editor

Any MCP Client

Streamable HTTP

Ready to connect?

Create an account, grab your API token, and add the MCP config. You'll be generating speech and video from your LLM in minutes.

Get Started Free API Documentation