Definition
Higgsfield MCP is Higgsfield AI's hosted connector built on the Model Context Protocol (MCP) — an open standard that lets AI agents securely access external tools. Higgsfield runs an MCP server (at mcp.higgsfield.ai) that exposes its entire media-generation stack: image and video models, character training, asset management, and virality scoring.
How it works
You connect your agent to the Higgsfield MCP server once. After that, you describe what you want in plain language and the agent handles model selection, parameter configuration, generation, and returning the result to the chat. There is no API key for the connector — you sign in with a Higgsfield account and usage is billed against your credits.
- Add the Higgsfield connector in your agent's settings.
- Paste the server URL:
https://mcp.higgsfield.ai/mcp. - Sign in with your Higgsfield account.
- Ask the agent to generate an image or video.
Which agents it supports
- Claude — web, desktop, mobile, and Claude Code (CLI recommended for Code).
- OpenClaw, Hermes Agent, NemoClaw.
- Cursor and Perplexity.
- Any MCP-compatible client.
What you can generate
The connector exposes 30+ image and video models through one connection, including Soul, Cinema Studio, Flux, Seedream, Kling, Veo, Nano Banana Pro, and GPT Image. Output runs up to 4K and any aspect ratio. Core capabilities include:
- Image and video generation from text, images, audio, or existing footage.
- Soul character training for consistent, reusable characters.
- Cinematic image-to-video animation with preset camera moves.
- Viral clip generation and virality prediction before publishing.
Common use cases
- Product launch videos from a product URL.
- E-commerce and social visuals at scale.
- Character-consistent campaigns and series.
- Quick comparisons across multiple image/video models.