🎬 Adobe Premiere Pro MCP Server

Control Adobe Premiere Pro with natural language using Claude, Cursor, and Antigravity.
Automate rough cuts, silence removal, B-roll placement, Lumetri grading, MOGRT titles, chapter markers, and batch exports via the Model Context Protocol.

Features • Architecture • 3-Minute Quickstart • Tool Catalog • AI Client Setup • AI Workflows • Troubleshooting

💡 What is PremiereProMCP?

PremiereProMCP is an open-source Model Context Protocol (MCP) server written in JavaScript (Node.js) that bridges modern LLMs directly to Adobe Premiere Pro.

Unlike DaVinci Resolve Studio (which exposes an external Python socket library), Premiere Pro's Document Object Model (DOM) is traditionally confined inside an ExtendScript / CEP runtime. PremiereProMCP solves this by establishing a bidirectional loopback bridge between your desktop AI assistant and Premiere Pro:

• 🗣️ Natural Language Control: "Inspect active sequence, split at 12.4s, ripple delete the silence, and add a green chapter marker."

• ⚡ Dual Execution Engines: Low-latency WebSocket bridge via an in-app CEP panel + zero-configuration macOS AppleScript fallback.

• 🛡️ Non-Destructive by Design: Includes sequence duplication safety tools before AI makes timeline modifications.

• ⏱️ Tick-Accurate Timing: Seamlessly converts seconds to Premiere Pro’s native internal time precision (254,016,000,000 ticks per second).

Related MCP server: Adobe Premiere Pro MCP

🌟 Key Features

🏗️ Architecture

flowchart LR
    subgraph AI Client
        A["Claude Desktop / Cursor / Antigravity"]
    end

    subgraph MCP Server
        B["PremiereProMCP Server<br/>(Node.js / stdio)"]
    end

    subgraph Premiere Pro Bridge
        C["CEP Panel (WebSocket)<br/>(ws://127.0.0.1:9098)"]
        D["macOS AppleScript<br/>(Zero-Install Fallback)"]
    end

    subgraph Adobe Premiere Pro
        E["Premiere Pro DOM<br/>(Sequences, Tracks, Clips, Lumetri)"]
    end

    A <-->|"MCP JSON-RPC (stdio)"| B
    B <-->|"WebSocket"| C
    B -.->|"osascript do script"| D
    C <-->|"ExtendScript JSX"| E
    D -.->|"ExtendScript JSX"| E

How the Bridge Works

1. The MCP Server (src/index.js) runs locally as a Node.js process using standard I/O (stdio).

2. The In-App Extension Panel (premiere-plugin/) runs inside Premiere Pro. On launch, it connects to the local WebSocket server (ws://127.0.0.1:9098).

3. When your AI assistant asks to perform an action (e.g. razor_clip), the server sends a structured JSON payload to the panel, which executes the corresponding Premiere Pro ExtendScript call and returns the result back to the LLM.

4. macOS Fallback: If the extension panel is not open, the server can automatically execute ExtendScript directly via AppleScript (osascript -e 'tell application "Adobe Premiere Pro" to do script ...').

🚀 3-Minute Quickstart

Step 1: Clone and Install Dependencies

git clone https://github.com/au-yong/premiere-pro-mcp.git
cd premiere-pro-mcp
npm install

Step 2: Install the Premiere Pro Extension Panel

Run the automated installer script:

npm run install-plugin

Step 3: Open the Extension in Premiere Pro

1. Launch Adobe Premiere Pro (2020 through 2025+).

2. Open any video project.

3. In the top application menu, select: Window ➔ Extensions ➔ Premiere Pro MCP Bridge

4. A sleek dark panel will open and show Connected to ws://127.0.0.1:9098.

Step 4: Test the Bridge Connection

Verify the end-to-end connection by running the test suite:

npm test

📦 Building with MCPB (Model Context Protocol Bundle)

PremiereProMCP supports the official MCPB (.mcpb) bundle specification for one-click installation and distribution in Claude Desktop and other MCPB-compliant clients without needing manual git clones or path configuration.

1. Validate the Bundle Manifest

npm run validate:mcpb

(Validates manifest.json against the latest MCPB v0.4 specification)

2. Pack the .mcpb Extension

npm run bundle:mcpb

This compiles and bundles the server, runtime dependencies, and manifest into a single portable package:

premiere-pro-mcp.mcpb

3. One-Click Installation

• Claude Desktop / MCPB Hosts: Drag and drop premiere-pro-mcp.mcpb directly into Claude Desktop or double-click to install.

• Configurable Settings: During installation, the host client will automatically prompt for optional configuration:

  • WebSocket Port (default: 9098)

  • Premiere App Name (default: Adobe Premiere Pro 2025)

🤖 AI Assistant Configuration

Configure your favorite AI assistant to launch PremiereProMCP via stdio.

Claude Desktop

Edit your Claude Desktop configuration file:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

• Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "premiere-pro": {
      "command": "node",
      "args": ["/ABSOLUTE/PATH/TO/PremiereProMCP/src/index.js"],
      "env": {
        "PORT": "9098",
        "HOST": "127.0.0.1"
      }
    }
  }
}

Cursor IDE

Create or update .cursor/mcp.json in your workspace or global settings:

{
  "mcpServers": {
    "premiere-pro": {
      "command": "node",
      "args": ["/ABSOLUTE/PATH/TO/PremiereProMCP/src/index.js"]
    }
  }
}

Google Antigravity / Gemini CLI

In your Antigravity or MCP client configuration:

{
  "mcpServers": {
    "premiere-pro": {
      "command": "node",
      "args": ["/ABSOLUTE/PATH/TO/PremiereProMCP/src/index.js"]
    }
  }
}

🛠️ Complete Tool Catalog

PremiereProMCP exposes over 30 tools organized across video editing domains:

1. Project & Media Pool Management

2. Sequence & Timeline Operations

3. Timeline Editing & Clip Manipulation

4. Motion, Transform & Keyframing (Inspector)

5. Effects & Color Grading (Lumetri)

6. Audio, Graphics & Markers

7. Export & Render

8. Compound AI Workflows & Utilities

⚡ High-Value Compound AI Workflows

Here is how you can use natural language to trigger compound video editing workflows:

1. The Auto-Silence Rough-Cutter

Prompt: "Analyze the audio on my active sequence. I want you to duplicate the sequence for safety, split clip at [04.2s, 08.5s, 14.1s], and ripple-delete the pauses."

1. AI calls duplicate_sequence(newName: "Rough_Cut_Backup").

2. AI calls batch_razor_cuts(...) at the silence boundary timestamps.

3. AI calls delete_clip(rippleDelete: true) to close the gaps into a punchy jump-cut.

2. YouTube Smart Chapter Generator

Prompt: "Based on this script outline, place chapter markers on the timeline: 00:00 Intro, 01:25 Setup, 04:10 Demo, 08:30 Conclusion."

• AI calls generate_youtube_chapters(...) which automatically calculates tick positions and drops named, color-coded markers directly onto the timeline.

3. Vertical Shorts / Reels 9:16 Reformatter

Prompt: "Create a 9:16 vertical version of my sequence for Instagram Reels, and scale the talking head by 150%."

1. AI calls reformat_aspect_ratio(aspectRatio: "9:16").

2. AI calls set_clip_transform(scale: 150, position: [540, 960]) to frame the presenter perfectly.

4. AI Lower-Third & Title Injector

Prompt: "Add the speaker lower-third 'Dr. Jane Smith - Lead AI Scientist' at 00:15 using my brand template."

1. AI calls import_mogrt(mogrtPath: "/templates/lower_third.mogrt", timeSeconds: 15.0).

2. AI calls update_mogrt_text(textValue: "Dr. Jane Smith\nLead AI Scientist").

🔍 Troubleshooting & FAQ

1. Ensure you ran npm run install-plugin.

2. Restart Adobe Premiere Pro completely (Premiere only scans the CEP/extensions folder at startup).

3. If on macOS, verify PlayerDebugMode was set:

   defaults read com.adobe.CSXS.11 PlayerDebugMode
   # Should output: 1

Create or update .env in the root of PremiereProMCP:

PORT=9100
HOST=127.0.0.1

And make sure the same URL (ws://127.0.0.1:9100) is entered in the panel settings input inside Premiere Pro.

PremiereProMCP supports Adobe Premiere Pro 2020 through 2025+ (internal versions 13.0 through 99.0). It runs on both macOS (Apple Silicon & Intel) and Windows 10/11.

Premiere Pro's standard public DOM does not expose a native clip.razor() method. To cut clips programmatically without third-party plugins, PremiereProMCP enables the QE (Quality Engineering) DOM (app.enableQE()), which is standard practice in professional Premiere Pro automation.

🛠️ Project Structure

PremiereProMCP/
├── src/
│   ├── index.js                  # Main MCP Server entry point (stdio)
│   ├── config.js                 # Environment config & constants
│   ├── bridge/
│   │   ├── bridge-manager.js     # Unified execution dispatcher
│   │   ├── websocket-bridge.js   # Real-time WebSocket server (ws://)
│   │   ├── applescript-bridge.js # macOS osascript fallback
│   │   ├── extendscript-helpers.js # JSON2 polyfill & tick math
│   │   └── extendscript-templates.js # DOM execution templates
│   ├── tools/                    # Modular MCP tools by category
│   │   ├── project-tools.js
│   │   ├── sequence-tools.js
│   │   ├── editing-tools.js
│   │   ├── inspector-tools.js
│   │   ├── effects-tools.js
│   │   ├── audio-tools.js
│   │   ├── graphics-tools.js
│   │   ├── marker-tools.js
│   │   ├── export-tools.js
│   │   ├── compound-tools.js
│   │   └── system-tools.js
│   └── resources/
│       └── index.js              # Live URI resources
├── premiere-plugin/              # CEP Extension Panel
│   ├── CSXS/manifest.xml         # Adobe CEP Extension manifest
│   ├── index.html                # Dark-themed status & log panel UI
│   ├── js/
│   │   ├── CSInterface.js        # Adobe CSInterface library
│   │   └── main.js               # WebSocket client & action runner
│   └── jsx/
│       └── hostscript.jsx        # ExtendScript host engine
├── scripts/
│   └── install-plugin.js         # One-command installer
├── test/
│   └── test-server.js            # Automated integration tests
└── package.json

🤝 Contributing

Contributions are very welcome! If you'd like to add new tools (such as multicam switching, transcript parsing, or Essential Sound auto-ducking):

1. Fork the repository.

2. Create a feature branch: git checkout -b feature/amazing-feature.

3. Test your changes with npm test.

4. Commit and open a Pull Request.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.