Runway just shipped an official MCP integration, letting any Model Context Protocol-compatible agent call Runway's image and video models directly. That includes Claude Desktop, Cursor, Codex, Cline, and the growing list of agents that have adopted MCP this year. Announced May 27, 2026, the integration removes the context-switch tax that has slowed agent-driven media workflows since the first wave of MCP releases.

How to integrate Runway MCP in your agent today

The deployment is straightforward. Add the Runway MCP server to your agent's MCP config (the exact path depends on the client: Claude Desktop uses claude_desktop_config.json, Cursor uses its Settings UI), restart the agent, and authenticate with your Runway account. Once connected, your agent can call Runway's image and video generation endpoints as tools, with the model selecting the right one based on your prompt.

A practical use case: ask Claude to draft a video storyboard, then have it call Runway through MCP to generate the actual shots for each beat, all in one conversation. Same flow works in Cursor for product demos or in any custom agent built on the protocol. The agent passes prompts, receives URLs, and embeds the output inline.

Why It Matters

MCP has become the default plumbing for agent-to-tool calls in 2026. Anthropic shipped it, OpenAI's Codex IDE consumes it, and now the creative-tool side is catching up. Runway joining the MCP ecosystem is the second major creator-tool integration this week after Krea 2 landed as a ComfyUI Partner Node and Mistral Vibe shipped its VS Code extension. Each release reduces the gap between thinking about a shot and getting pixels on screen.

For working creators, this changes the production loop. You no longer leave your editor to open runwayml.com, paste a prompt, wait, download, and re-import. The model becomes a callable function inside whatever surface you already work in.

Key Details

The MCP server exposes Runway's generation API surface for both image and video endpoints documented in the Runway API reference. Authentication uses your existing Runway account credit balance, so an MCP-driven generation pulls from the same pool as web or API calls. There's no separate billing tier and no separate quota.

Runway has not published a separate developer-tier price for MCP usage. Calls are billed at the same per-second video and per-image rates as the public API. The integration is opt-in: agents only see Runway tools if you've configured the MCP server in your client, following the same pattern as other recent MCP-adjacent voice and creator integrations.

What to Do Next

Install the Runway MCP server in whichever agent you use most, generate one shot end-to-end to confirm the connection, then look at which step in your current production workflow could be collapsed into the agent. If you build any video flow that involves storyboarding plus shot generation, this is the integration that closes the loop.