Mographr MCP Server

Mographr's MCP server lets an AI assistant render real motion-graphics overlays — animated lower thirds, titles, callouts — and hand back a finished MP4. It is free, anonymous and needs no API key: point any MCP client at https://mcp.mographr.com/mcp. Unlike text-to-video models, it renders designed, typographically exact graphics you drop over your own footage.

What this is — and what it isn't

This is not a text-to-video model. It does not invent footage. It renders a small, hand-built catalogue of motion-graphics templates with your text, your colour scheme and your choice of design, and returns a real MP4 you composite over your own video. A diffusion model asked for "a lower third that says Dr. Maya Chen" produces something that looks like a lower third and spells the name wrong. This renders actual text in an actual font on an actual keyframed animation, so it is legible at 1080p, identical on every run, and safe to put in front of a client.

How do I connect an MCP client to Mographr?

Claude Code

claude mcp add --transport http mographr https://mcp.mographr.com/mcp

Claude Desktop

Settings → Connectors → Add custom connector. Older builds without that UI use claude_desktop_config.json with the mcp-remote bridge:

{
  "mcpServers": {
    "mographr": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote@latest",
        "https://mcp.mographr.com/mcp"
      ]
    }
  }
}

Cursor

~/.cursor/mcp.json, or .cursor/mcp.json inside a project:

{
  "mcpServers": {
    "mographr": {
      "type": "http",
      "url": "https://mcp.mographr.com/mcp"
    }
  }
}

VS Code

.vscode/mcp.json:

{
  "servers": {
    "mographr": {
      "type": "http",
      "url": "https://mcp.mographr.com/mcp"
    }
  }
}

Anything else

It is plain HTTP POST — no SDK required:

curl -sS https://mcp.mographr.com/mcp \
  -H 'Content-Type: application/json' \
  -H 'Accept: application/json, text/event-stream' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Gotcha: The Accept header must list BOTH application/json and text/event-stream — the Streamable HTTP transport rejects the request otherwise. This trips up almost every hand-rolled client.

What tools does the server expose?

What are the limits?

Frequently asked questions

Do I need an API key or an account?

No. The server is deliberately anonymous — there is no key to obtain, no signup, and no billing hook. Point your MCP client at the endpoint and call the tools. The trade-off is rate limiting by IP address rather than by account, and a watermark on the free MP4. If you want unwatermarked or transparent-background output, that lives in the Mographr studio on a paid plan, but nothing about the MCP server requires it.

How is this different from a text-to-video MCP server?

Text-to-video servers generate footage from a prompt: every run differs, and text inside the frame is hallucinated letterforms rather than real type. This renders a designed template — real fonts, real kerning, exact keyframes — with the text you supply. The same input always produces the same file. Use a text-to-video server when you need a shot that does not exist; use this when you need a name tag, title card or callout to composite over footage you already have.

What does it actually return?

A URL to a finished MP4, roughly 30 to 70 seconds after you call render_motion_tool. The flow is: render_motion_tool gives you a renderId, then you poll get_motion_render until status is "completed" and read downloadUrl. The file is a normal H.264 MP4 at the design's native canvas size, ready to drop on a timeline in Premiere, Resolve, Final Cut or CapCut.

Can it render vertical video for Reels, Shorts or TikTok?

Yes. Several designs are authored at 1080x1920 rather than 1920x1080, and each variant reports its own width and height from get_motion_tool. You do not pass a size — the design carries its canvas, so pick a vertical variant and the render comes back vertical.

Is the output free to use commercially?

You own the output generated for you, subject to the Terms. Note that AI-assisted output carries no guarantee of copyright protection in every jurisdiction, which is a property of the law rather than of this service. The free render includes a small mographr.com watermark; removing it is a paid upgrade.

What happens when I hit the rate limit?

The tool returns a normal, successful result whose payload says status "rate_limited", how many seconds to wait, and a URL where the same render can be produced immediately with no limit. It deliberately does not return an MCP error, because assistants surface errors as "this tool is broken" — being asked to wait is not a failure, and the distinction matters when someone is evaluating the server for the first time.

Is there a REST API as well?

Yes, and it is the same engine. GET https://bff.mographr.com/tools lists the catalogue, GET https://bff.mographr.com/tools/<slug> returns the schema, and POST https://bff.mographr.com/tools/<slug>/render starts a render. REST also supports image upload, which MCP does not. Use REST if you are writing a normal backend integration; use MCP if you want an assistant to drive it conversationally.

Does the server keep a session?

No. The transport is stateless Streamable HTTP: every POST is independent, there is no session id header and nothing to clean up. A deploy can therefore never strand your connection. GET and DELETE on the endpoint answer 405 by design, and CORS is open, so browser-based clients work too.

See also