MakeAIVideo

MakeAIVideo MCP Server: AI video generation for ChatGPT, Claude, Cursor and any AI agent

The MakeAIVideo MCP server lets ChatGPT, Claude, Cursor and any AI agent create short-form AI videos (script, AI voiceover, AI or stock scenes, captions and music) from a prompt or a script, return the finished MP4, and post or schedule it to TikTok, Instagram Reels, YouTube Shorts and other connected social accounts.

It is the official Model Context Protocol server for MakeAIVideo, an AI video generator for short-form video.

What you can make

Ask your assistant for a video in plain words and it picks the right tool:

• TikTok, Instagram Reels and YouTube Shorts videos from one prompt: the script, voiceover, scenes, captions and music are made for you (TikTok video generator, Instagram Reels generator, AI YouTube Shorts generator).
• Videos from your own script, narrated as written (script to video), or from a blog post or article URL (blog to video).
• Explainers, list videos and story videos for faceless channels (AI explainer video, faceless YouTube channel).
• Talking avatar and presenter videos: a lip-synced presenter reads your message straight to camera (talking avatar, AI spokesperson video).
• Product demos, UGC-style ads and video ads for Meta and TikTok (AI UGC video, AI ad maker).
• Captions and music: captions on every video, captions as SRT or VTT files, AI background music made from a prompt, and your own choice of voice.
• Posting and scheduling to TikTok, Instagram, YouTube, Facebook Pages, LinkedIn, X, Threads, Pinterest, Bluesky, Telegram and Discord (auto-post).

Image to video and character swap need a file upload, so they run in the web app and the REST API, not through MCP tools. See more on the examples page.

Example prompts

• "Make a 30-second vertical explainer video about why octopuses have three hearts and give me the MP4."
• "Give me 10 video ideas for a personal finance TikTok channel, then make the best one."
• "Write a script for a UGC-style ad for my skincare serum. Show me the script before you make the video."
• "Turn the blog post at this link into a 60-second YouTube Short." (then paste the URL)
• "Make a 20-second spokesperson video announcing our new opening hours, with a calm British voice."
• "How many credits would a 45-second list video cost?"
• "When the video is ready, post it to my TikTok and YouTube tomorrow at 9am with a short caption."

Quick start

The hosted server is the easiest way in. Add https://mcp.makeaivideo.ai, sign in to MakeAIVideo in the browser window that opens, and press Approve. You do not paste a key. The steps for every app, with copy buttons, are on the AI Assistants page in the app.

After you connect, try: *"How many MakeAIVideo credits do I have? Then give me three video ideas for my channel."*

Claude (claude.ai and Claude Desktop)

1. In Claude, open Customize → Connectors, press +, then Add custom connector.
2. Name it MakeAIVideo and paste https://mcp.makeaivideo.ai.
3. Press Add, then Connect. Sign in to MakeAIVideo and press Approve.

It then works in the Claude phone apps too. On Team and Enterprise plans an Owner adds connectors first, in Organization settings.

ChatGPT

1. On the web, switch on Developer mode: Settings → Apps → Advanced settings. Custom connectors stay hidden until it is on.
2. In Settings → Apps, choose Create. Name it MakeAIVideo and paste https://mcp.makeaivideo.ai.
3. Set authentication to OAuth, press Create, then sign in to MakeAIVideo and approve.

Developer mode needs a Plus, Pro, Business, Enterprise or Education plan.

Claude Code

claude mcp add --scope user --transport http makeaivideo https://mcp.makeaivideo.ai

Then type /mcp in Claude Code, choose Authenticate, and sign in. Keep --scope user so the server works in every folder.

Cursor

Click Add to Cursor above and confirm Install, or add this to ~/.cursor/mcp.json:

{

"mcpServers": { "makeaivideo": { "url": "https://mcp.makeaivideo.ai" } } }

Approve the sign-in window Cursor opens.

VS Code (GitHub Copilot agent mode)

Click Install in VS Code above, or run MCP: Add Server from the Command Palette, choose HTTP and paste the address. Or add this to .vscode/mcp.json:

{

"servers": { "makeaivideo": { "type": "http", "url": "https://mcp.makeaivideo.ai" } } }

Sign in when VS Code asks, then open Chat in Agent mode.

Codex

codex mcp add makeaivideo --url https://mcp.makeaivideo.ai

codex mcp login makeaivideo

Editing ~/.codex/config.toml by hand? The header is [mcp_servers.makeaivideo], with an underscore.

Gemini CLI

gemini mcp add --scope user --transport http makeaivideo https://mcp.makeaivideo.ai

Start Gemini CLI, run /mcp auth makeaivideo, then sign in. Or install the MakeAIVideo extension from this repo:

gemini extensions install https://github.com/makeaivideo-ai/mcp

Windsurf

Add this to ~/.codeium/windsurf/mcp_config.json, then fully quit and reopen Windsurf and approve the sign-in:

{

"mcpServers": { "makeaivideo": { "serverUrl": "https://mcp.makeaivideo.ai" } } }

Other apps

Agents that sign themselves in can follow auth.md (OAuth with PKCE, or the device code flow). Prefer an API key? Send Authorization: Bearer mav_... and create the key in the app under Developers.

Local server (npx)

For apps that only run local commands. The package is a small stdio bridge to the hosted server, so the tools are always the same as the hosted ones. Node 18 or later.

{

"mcpServers": { "makeaivideo": { "command": "npx", "args": ["-y", "@makeaivideo/mcp"], "env": { "MAKEAIVIDEO_API_KEY": "mav_YOUR_KEY" } } } }

Works with Claude Desktop, Cursor, Windsurf, Cline, Zed, Continue and any stdio MCP client.

With Docker (the Dockerfile in this repo):

docker build -t makeaivideo-mcp .

docker run -i --rm -e MAKEAIVIDEO_API_KEY=mav_YOUR_KEY makeaivideo-mcp

Without a key the local server still starts and lists its tools from the bundled tools.json, so clients can inspect it. Every tool call then asks for a key.

Tools (44)

Create videos

Each create tool takes the length, aspect ratio (9:16, 1:1 or 16:9), and an optional character and voice. It returns a video id at once; the render takes minutes.

Plan and write

Track and deliver

Edit and redo

| regenerate_voice | Re-records the