Runway API MCP Server

Instructions

Implementation Reference

• src/index.ts:141-152 (registration)

  Registration of the 'runway_upscaleVideo' tool using server.tool(), specifying name, description, input schema, and handler function.

  server.tool(
    "runway_upscaleVideo",
    "Upscale a video to a higher resolution. videoUri takes in a url of a video or a data uri of a video.",
    { videoUri: z.string() },
    async ({ videoUri }) => {
      const task = await callRunwayAsync("/video_upscale", {
        method: "POST",
        body: JSON.stringify({ videoUri, model: "upscale_v1" }),
      });
      return { content: [{ type: "text", text: JSON.stringify(task) }] };
    }
  );

• src/index.ts:145-151 (handler)

  The handler function that executes the tool: sends POST to /video_upscale endpoint with videoUri and model, awaits task completion, returns JSON stringified task.

  async ({ videoUri }) => {
    const task = await callRunwayAsync("/video_upscale", {
      method: "POST",
      body: JSON.stringify({ videoUri, model: "upscale_v1" }),
    });
    return { content: [{ type: "text", text: JSON.stringify(task) }] };
  }

• src/index.ts:144-144 (schema)

  Input schema using Zod: requires 'videoUri' as a string (URL or data URI of the video).

  { videoUri: z.string() },

• src/index.ts:59-72 (helper)

  Helper function used by the handler to call Runway API and automatically wait for asynchronous task completion by polling.

  async function callRunwayAsync(
    path: string,
    opts: Partial<RequestInit> = {}
  ): Promise<RunwayTask> {
    const response = (await callRunway(path, opts)) as {
      id?: string;
    } & RunwayTask;
    // If the response has a taskId, wait for completion
    if (response?.id) {
      return waitForTaskCompletion(response.id);
    }
    // If no taskId, just return the response as is
    return response;
  }