Chapter 11: External AI Agent Integration (MCP)
Connect Claude Code, Cursor, and Antigravity to Scribis via Model Context Protocol (MCP) for automated editing.
Chapter 11: External AI Agent Integration (MCP)
11.1 What is Scribis MCP Integration?
Scribis natively implements the open Model Context Protocol (MCP) standard.
Scribis MCP integration diagram showing external AI agents (Claude Code / Cursor / Antigravity) commanding Scribis desktop
MCP gives external AI agents (such as Claude Code, Cursor, Google Antigravity IDE, and Windsurf) direct access to:
- 👂 Read Subtitles: AI can read live project transcripts.
- 👁️ Visual Understanding: AI inspects video frames at specific timestamps.
- ✂️ Timeline Automation: AI marks rough cut deletions, applies zoom-in pans, and adds privacy mosaics.
- 🗣️ Voice Synthesis: AI triggers Kokoro TTS to generate voiceovers.
11.2 Enabling MCP in Scribis
Scribis MCP Protocol settings screen showing service toggle, HTTP port 8890, and copy config button
- In Scribis, navigate to "Settings -> MCP Protocol".
- Toggle on "Enable MCP Service".
- Copy the generated configuration snippet into your AI client's
mcp.json.
11.3 Popular AI Agent Prompts
Once configured, instruct your AI agent in natural language:
Prompt 1: "AI Subtitle Proofreader"
"Read
scribis://subtitles/current, identify any homophone errors or missing punctuation, and submit corrections back to Scribis for review."
Prompt 2: "Automated Privacy Mosaic"
"Check the video frames between 00:15 and 00:30. If any license plates or ID cards are visible, add a mosaic block over those coordinates."
Prompt 3: "Highlight Reel & Dynamic Zoom"
"Analyze the transcript to find the top 3 most exciting moments. Mark the surrounding silence as rough-cut deletions and apply a 1.4x zoom-in to the key statements."