Connections & Capabilities
Connects To
Capabilities
Quickstart
Install
pip install -rConfig
{
"mcpServers": {
"davinci-resolve": {
"name": "DaVinci Resolve MCP",
"command": "/path/to/your/venv/bin/python",
"args": [
"/path/to/your/davinci-resolve-mcp/src/main.py"
]
}
}
}Exposed MCP Tools (6)
Get DaVinci Resolve versionRetrieves the version number of the DaVinci Resolve application.
Read-only operation with no side effects.
Create new timelineCreates a new timeline within the current DaVinci Resolve project.
Non-destructive write operation that adds a new element to the project.
Import media fileImports a specified media file into the DaVinci Resolve media pool.
Adds a new media file to the project, but does not modify existing files.
Open project by nameOpens a DaVinci Resolve project specified by its name.
Changes the current project, but does not modify project files directly.
Add clip to timelineAdds a specified clip from the media pool to the current timeline.
Modifies the timeline, but does not affect the underlying media files.
Save current projectSaves the current DaVinci Resolve project.
Writes changes to the project file, but does not delete or overwrite other files.
Safety Assessment
The MCP server offers convenient control over DaVinci Resolve but introduces risks due to write capabilities. It is safe when used with careful command construction and a clear understanding of the potential impact on DaVinci Resolve projects. Risky scenarios involve automated or unsupervised operation where unintended modifications can occur.
- Requires DaVinci Resolve to be running, limiting scope to the application.
- Client-specific launch scripts provide some isolation.
- No direct internet exposure from the MCP server itself.
- Clear separation of concerns between the MCP server and DaVinci Resolve.
- Allows write operations within DaVinci Resolve, potentially modifying projects.
- Relies on environment variables, which can be a source of misconfiguration.
- No explicit RBAC or permission controls within the MCP server.
- Potential for unintended actions if natural language commands are misinterpreted.
