MCP: letting Claude edit your videos

Most products that say “works with AI” mean a chatbot widget bolted onto the corner of the page. We mean something more literal: EditorBuddy runs an MCP (Model Context Protocol) server, so Claude — or any MCP-compatible agent — can plan, edit, and publish a video the same way the web app does.

What that looks like in practice

Add EditorBuddy as an MCP server and Claude gets the same toolset our orchestrator uses internally: transcribe, cut, caption, translate, dub, find clips, publish. Ask Claude to “take this recording, cut the best three minutes, caption it, and post it to my channels,” and it can call those tools directly — no copy-paste between apps, no exporting a file just to re-upload it somewhere else.

{
  "mcpServers": {
    "editorbuddy": {
      "url": "https://mcp.editorbuddy.com"
    }
  }
}

Same brain, same memory, same limits

This isn’t a stripped-down API surface for external agents. A job kicked off through MCP runs through the exact same orchestrator DAG, reads from the same memory of your style, and respects the same per-job cost caps as a job started from the web app. From the orchestrator’s point of view, “a request from Claude” and “a request from the EditorBuddy UI” are the same kind of thing.

We didn’t build an API and call it agent-friendly. We built the agent first, and the web app is one of its clients.

Progress you can narrate

Long video jobs take time, so the MCP server streams progress the same way the web app does — over SSE. A calling agent can tell its user “transcription done, cutting clips now” instead of going silent for two minutes and hoping.

Where this goes

The web app will always be the fastest way for a person to sit down and edit a video by hand. But increasingly, the request to “edit a video” doesn’t start with a person opening a browser — it starts with a sentence to an assistant that already has the file. We want EditorBuddy to be the thing that assistant reaches for.