Resolve MCP
README.md
# Resolve MCP
A production-ready Model Context Protocol (MCP) server that connects AI assistants like Claude to DaVinci Resolve Studio for professional video editing automation.
## Features
- **Project Inspection**: Get project info, timelines, tracks, clips, and media pool contents
- **Media Import**: Import media files into the media pool with folder organization
- **Timeline Editing**: Cut, trim, ripple delete, add markers, and Text+ titles
- **Clip Properties**: Adjust opacity, volume, transform, speed, and color grade
- **Timeline Management**: Duplicate timelines, export to EDL/FCPXML/OTIO
- **Visual Verification**: Render frames and take screenshots for AI visual verification
- **Safety First**: Approval system for destructive actions
- **Cross-Platform**: macOS, Windows, and Linux support
- **MCP Compatible**: Works with Claude Desktop and other MCP clients
## Installation
### Via npx (Recommended)
```bash
npx resolve-mcp
```
### Global Install
```bash
npm install -g resolve-mcp
resolve-mcp
```
### Claude Desktop Configuration
Add to your Claude Desktop config (`~/Library/Application Support/Claude/claude_desktop_config.json` on macOS):
```json
{
"mcpServers": {
"resolve-mcp": {
"command": "npx",
"args": ["resolve-mcp"],
"env": {}
}
}
}
```
## Requirements
- **Node.js** 20+
- **Python** 3.8+ (for the Resolve bridge)
- **DaVinci Resolve Studio** 18+ (Free version has limited scripting support)
- **DaVinci Resolve running** with a project open
## Usage
### Start the MCP Server
```bash
# Default (stdio transport for MCP clients)
npx resolve-mcp
# Test connection
npx resolve-mcp test
# System check
npx resolve-mcp doctor
# With auto-approve (testing only)
npx resolve-mcp serve --auto-approve
```
### Available Tools
Once connected to an MCP client (like Claude), you can use these tools:
#### Project & Timeline Inspection
- `get_project_info` - Get full project details
- `get_timeline_info` - Get specific timeline details
- `list_timelines` - List all timelines
- `get_media_pool` - Get media pool contents
- `search_media_pool` - Search for clips
#### Media Import
- `import_media` - Import media files
#### Editing Operations
- `cut_clip` - Cut a clip at a specific frame
- `trim_clip` - Trim clip start/end with optional ripple
- `ripple_delete` - Remove timeline section with ripple edit
- `add_marker` - Add markers to timeline
- `add_text_title` - Add Text+ titles
- `duplicate_timeline` - Duplicate a timeline
- `adjust_clip_properties` - Adjust clip properties
#### Export
- `export_timeline` - Export to EDL, FCPXML, or OTIO
#### Visual Verification
- `render_frame` - Render a single frame
- `take_screenshot` - Take timeline viewer screenshot
- `get_resolve_version` - Get Resolve version info
## Example Workflows
### Basic Editing with Claude
```
User: "Open my project and show me the timeline structure"
Claude: [Calls get_project_info, lists timelines and tracks]
User: "Cut the clip at 00:01:30 and add a marker called 'Scene Change'"
Claude: [Calls cut_clip and add_marker with approval]
User: "Export the timeline as FCPXML for Final Cut Pro"
Claude: [Calls export_timeline with format: fcpXML]
```
### Visual Verification
```
User: "Show me frame 1500 of the timeline to verify the cut"
Claude: [Calls render_frame, returns image for visual verification]
User: "Take a screenshot at 00:02:15:10"
Claude: [Calls take_screenshot with timecode]
```
## Safety & Approval System
All destructive operations require explicit approval:
- **High Risk**: Ripple delete, trim with ripple, color grade changes
- **Medium Risk**: Cut, trim, markers, titles, duplicate, property adjustments, export, import
- **Low Risk**: Read-only operations (project info, timeline info, media pool)
The approval system integrates with MCP clients to prompt users before executing risky operations.
## Architecture
```
┌─────────────────┐ MCP Protocol ┌──────────────────┐
│ AI Assistant │ ◄────────────────────► │ Resolve MCP │
│ (Claude, etc.) │ stdio/JSON-RPC │ Server (Node.js)│
└─────────────────┘ └────────┬─────────┘
│
spawn/stdin
▼
┌──────────────────────┐
│ Python Bridge │
│ (resolve_bridge.py) │
└──────────┬───────────┘
│
Resolve Scripting API
▼
┌──────────────────────┐
│ DaVinci Resolve │
│ Studio │
└──────────────────────┘
```
## Development
```bash
# Install dependencies
npm install
# Build
npm run build
# Development mode with hot reload
npm run dev
# Run tests
npm test
# Lint
npm run lint
```
## Project Structure
```
resolve-mcp/
├── src/
│ ├── index.ts # Main exports
│ ├── cli.ts # CLI entry point
│ ├── types/index.ts # TypeScript types
│ ├── server/mcp-server.ts # MCP server implementation
│ ├── bridge/bridge-client.ts # Python bridge client
│ └── utils/approval.ts # Approval system
├── bridge/
│ └── resolve_bridge.py # Python Resolve bridge
├── tests/ # Test files
├── dist/ # Compiled output
├── package.json
├── tsconfig.json
└── README.md
```
## DaVinci Resolve Scripting Setup
### macOS
The scripting module is installed with Resolve at:
```
/Library/Application Support/Blackmagic Design/DaVinci Resolve/Developer/Scripting/Modules
```
### Windows
```
C:\ProgramData\Blackmagic Design\DaVinci Resolve\Support\Developer\Scripting\Modules
```
### Linux
```
/opt/resolve/Developer/Scripting/Modules
```
Ensure the `DaVinciResolveScript.py` module is in your Python path, or the bridge will auto-detect it.
## Troubleshooting
### "Could not connect to DaVinci Resolve"
- Ensure DaVinci Resolve is running
- Ensure a project is open
- Check that the scripting module is accessible
### "DaVinci Resolve scripting module not found"
- Verify Resolve Studio is installed (Free version has limited API)
- Check the platform-specific paths above
- Try running `python3 -c "import DaVinciResolveScript"`
### Permission denied on bridge script
```bash
chmod +x bridge/resolve_bridge.py
```
## License
MIT License - see LICENSE file for details.
## Contributing
1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Run tests and linting
5. Submit a pull request
## Support
- Issues: GitHub Issues
- Documentation: This README
- MCP Spec: https://modelcontextprotocol.ioThis server cannot be deployed
Maintenance
ActivityMaintained
ResponsivenessNo recent community issues