Modelrunner

What is Modelrunner?

Modelrunner is the official Model Context Protocol (MCP) server for the ModelRunner platform. It lets you run over 100 AI models across image, video, audio, and 3D generation directly from Claude, Cursor, and any MCP client using a single remote endpoint with OAuth 2.1 authentication.

How to use Modelrunner?

Add the server via claude mcp add --transport http modelrunner https://mcp.modelrunner.run/mcp in Claude Code, or set its URL in Cursor/VS Code’s MCP JSON config. For stdio‑only clients like Claude Desktop, use npx -y mcp-remote https://mcp.modelrunner.run/mcp. On first use your AI assistant opens a browser window to sign in — no API key pasting needed.

Key features of Modelrunner

• One endpoint exposes the entire ModelRunner catalog (100+ models)
• Pay‑per‑use billing with no subscription
• OAuth 2.1 — no API key required in config
• Tools for discovering, inspecting, and running models
• “Wrapper” system for creating reusable simplified model interfaces
• Fully remote server — nothing to install or self‑host

Use cases of Modelrunner

• Generate images, videos, music, or 3D models from conversational AI assistants
• Automate media generation pipelines without managing multiple API keys
• Search and compare models from different providers in one place
• Build custom wrappers with simplified inputs and prompt templates

FAQ from Modelrunner

How is Modelrunner different from single‑vendor MCP servers?

Unlike single‑vendor servers, Modelrunner gives access to models from many AI providers behind one endpoint and one pay‑per‑use billing account — no separate API keys or subscriptions.

Do I need to install anything to use Modelrunner?

No. The server is first‑party and hosted remotely by ModelRunner. The configuration snippets in the README are the entire integration; there is nothing to install or self‑host.

Where are the model outputs stored?

Inference results are returned as hosted URLs directly in the conversation. No local file storage is required.

How does authentication work?

Authentication uses OAuth 2.1. On first use your browser opens a ModelRunner sign‑in page; after authorizing, no API key is needed in the MCP configuration.

What is the pricing model?

Connecting the server is free. Model runs are pay‑per‑use from prepaid credits — no subscription. Each model’s exact price (per image, per second of video, etc.) is shown on its model page and on the ModelRunner pricing page.