mcp.film

mcp.film / Video Generation / MakeAIVideo MCP

MakeAIVideo MCP

OfficialRemoteby MakeAIVideocreditsverified Oct 18, 2026

Brief or script in, narrated and captioned MP4 out, over one OAuth endpoint.

What it does

MakeAIVideo's official hosted MCP turns a one-line brief, a supplied script, or a still image into a finished video — script, neural voiceover, generated scenes, caption timing, music bed and MP4 export in a single pass. The docs now enumerate 44 tools: format-specific creators (explainer, listicle, story, UGC, demo, article, spokesperson, from-script), planning and cost estimates, scene and voice regeneration, captions and share links, characters, voices and music, social publishing, and webhooks. For a filmmaker-agent this is the assembly end of the pipeline: explainers, spokesperson cuts and vertical social edits where finishing speed matters more than shot-level control.

Connect

Remote endpoint (Streamable HTTP)
https://mcp.makeaivideo.ai
Claude Code
claude mcp add --scope user --transport http makeaivideo https://mcp.makeaivideo.ai
Claude Desktop (claude_desktop_config.json)
{
  "mcpServers": {
    "makeaivideo": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.makeaivideo.ai"
      ]
    }
  }
}
Cursor (.cursor/mcp.json)
{
  "mcpServers": {
    "makeaivideo": {
      "url": "https://mcp.makeaivideo.ai"
    }
  }
}
Auth: oauth — Browser approval during connector setup (dynamic client registration + PKCE), or a mav_* key from Settings → Advanced → API keys

Sample tools

create_explainer_videocreate_video_from_scriptestimate_video_costregenerate_sceneget_video_statusget_download_urlpublish_videoget_credits

Field notes

Re-checked 18 Oct 2026: the docs (still headed v1.2) now list 44 tools rather than the 39 the endpoint advertised at launch, adding format-specific creators and social posting (list_social_accounts, publish_video). The older app.makeaivideo.ai/api/mcp address still works. Six weeks of history is not a maintenance record — re-verify before you build a paid pipeline on it. The vendor names no company, no jurisdiction and no repo, and describes its engines only as 'a curated stack of leading AI models', so you cannot audit which model renders your footage or what the output licence is. Unauthenticated calls return JSON-RPC error -32001 rather than an HTTP 401, which trips clients that key off status codes. The docs' own Claude Code example passes --header "Authorization: Bearer $MAV_API_KEY"; the command here omits it and relies on OAuth (read/write/ai scopes), so add the header for headless clients. An assistant spends the same credits you do: 7-day trial with 350 credits, then plans from $9/month. Overlaps heavily with invideo: both are templated script-to-video assemblers, and invideo costs nothing to try while this one has a real endpoint and real auth.

Also in Video Generation

Pairs well with