← marketplace
aitoolsha:1235724059627b5cmanual
runway-mcp
Run Runway's Gen-3 and Gen-4 video models from an agent — text-to-video, image-to-video, and upscaling — through Runway's official MCP server.
Install confidence
curl --create-dirs -fsSL https://skillmake.xyz/i/runway-mcp -o ~/.claude/skills/runway-mcp/SKILL.md
Pinned content
sha:1235724059627b5c
Generated with
manual
Source
github.com
The file served at /api/marketplace/runway-mcp-12357240/raw matches this hash. Inspect before install, then copy the command.
2,901 chars · ~725 tokens
--- name: runway-mcp description: "Run Runway's Gen-3 and Gen-4 video models from an agent — text-to-video, image-to-video, and upscaling — through Runway's official MCP server." source: https://github.com/runwayml/runway-api-mcp-server generated: 2026-05-28T19:59:37.331Z category: tool audience: ai --- ## When to use - Generating video from a text prompt using Runway Gen-3 Alpha or Gen-4 Turbo - Animating a reference image into video with image-to-video generation - Upscaling an existing video to higher resolution through the agent - Polling task status and retrieving finished video URLs without leaving the agent loop - Building an automated creative pipeline that produces video assets programmatically ## Key concepts ### Task-based generation Runway generation is async — create a task, get a task ID, poll until complete, then retrieve the output URL. The MCP handles all three steps as separate tool calls. ### 24-hour asset expiry Generated video URLs expire 24 hours after creation. Download and store assets before referencing them in downstream steps. ### Model versions Gen-3 Alpha is stable and widely supported; Gen-4 Turbo offers faster generation and higher fidelity. Specify the model when creating a task or default to the latest available. ### RUNWAYML_API_SECRET All API calls require a Runway API secret. Store it in env and never paste or log the value in chat or generated files. ## API reference ``` git clone https://github.com/runwayml/runway-api-mcp-server && npm install && npm run build ``` Build the MCP server locally — no npm package is published, so clone and build before registering with the MCP client. ``` git clone https://github.com/runwayml/runway-api-mcp-server cd runway-api-mcp-server npm install && npm run build # Then register: claude mcp add runway \ -e RUNWAYML_API_SECRET=key_xxx \ -- node /path/to/runway-api-mcp-server/build/index.js ``` ``` text_to_video / image_to_video / upscale_video / get_task tools ``` Use the Runway MCP tools through the client after registration. Always follow create → poll → retrieve pattern. ``` Ask the agent: "Use Runway to generate a 5-second cinematic video of a futuristic city at night, then return the download URL." ``` ## Gotchas - No npm package — must clone the repo and run npm run build before registering - Video URLs expire after 24 hours; save outputs immediately in pipelines that run across days - Generation is async; do not assume the video is ready in the same turn as the create call — poll until status is complete - Do not expose RUNWAYML_API_SECRET in logs, screenshots, or generated artifacts - Image-to-video input images must be accessible URLs or base64-encoded data — local file paths will not work --- Generated by SkillMake from https://github.com/runwayml/runway-api-mcp-server on 2026-05-28T19:59:37.331Z. Verify against source before relying on details.
File: ~/.claude/skills/runway-mcp/SKILL.md