Davinc-Resolve-Local-Bridge-MCP

Control DaVinci Resolve from any AI assistant — Claude, Cursor, Devin, and more.

218 tools. Resolve API and optional-extension coverage. Works with DaVinci Resolve Free and Studio.

What Is This?

Davinc-Resolve-Local-Bridge-MCP is a Model Context Protocol (MCP) server that gives AI assistants controlled access to DaVinci Resolve. It exposes 218 tools covering projects, timelines, media, Fusion, color grading, rendering, and supported optional extensions.

How It Works

Your AI Assistant (Claude, Cursor, Devin, etc.)
       │
       │  talks MCP (stdio)
       ▼
  mcp_server/server.py        ← runs on your machine
       │
       │  talks HTTP (localhost:8787)
       ▼
  DaVinciResolveBridge.py ← runs INSIDE DaVinci Resolve
       │                        (started from the Workflow Integration)
       ▼
  DaVinci Resolve API         ← full read + write access

The bridge script runs inside DaVinci Resolve's process, giving it full access to the Resolve scripting API — even on the Free edition (which blocks external scripting). The MCP server runs as a separate process and proxies tool calls to the bridge over localhost HTTP.

Quick Start

Prerequisites

• DaVinci Resolve 18.5+ (Free or Studio)
• Python 3.10+ on your machine
• An MCP-compatible AI assistant (Claude Desktop, Cursor, Devin, etc.)

Step 1 — Install the Workflow Integration and Bridge

Run the installer:

python install.py

The installer adds the user-facing launcher under Workspace → Workflow Integrations and installs the underlying bridge script. Restart Resolve once after installation so it discovers the integration.

For a manual installation, copy workflow/DaVinci Resolve Local Bridge MCP Server Connection.py and the workflow/assets directory to Resolve's Workflow Integration Plugins directory. Also copy bridge/Bridge-Connection-Script.py to Resolve's scripts folder as DaVinciResolveBridge.py. The Scripts entry is retained as a fallback, not the normal launch path.

Step 2 — Install MCP Server Dependencies

pip install -r requirements.txt

Step 3 — Configure Your AI Assistant

Add the MCP server to your assistant's configuration file. The server path is mcp_server/server.py in this repo.

Claude Desktop

Edit claude_desktop_config.json:

{
  "mcpServers": {
    "davinci-resolve-local-bridge": {
      "command": "python",
      "args": ["C:/path/to/Davinc-Resolve-Local-Bridge-MCP/mcp_server/server.py"]
    }
  }
}

Cursor

Edit .cursor/mcp.json:

{
  "mcpServers": {
    "davinci-resolve-local-bridge": {
      "command": "python",
      "args": ["C:/path/to/Davinc-Resolve-Local-Bridge-MCP/mcp_server/server.py"]
    }
  }
}

Devin

Edit mcp_config.json:

{
  "mcpServers": {
    "resolve-local-bridge": {
      "command": "python",
      "args": ["C:/path/to/Davinc-Resolve-Local-Bridge-MCP/mcp_server/server.py"]
    }
  }
}

Claude Code

claude mcp add davinci-resolve-local-bridge -- python C:/path/to/Davinc-Resolve-Local-Bridge-MCP/mcp_server/server.py

Any MCP Client (generic)

The server uses stdio transport. Launch it with:

python mcp_server/server.py

Step 4 — Start the Bridge Inside DaVinci Resolve

1. Open DaVinci Resolve
2. Go to Workspace → Workflow Integrations → DaVinci Resolve Local Bridge MCP Server Connection
3. Click Start Bridge
4. Confirm that the panel reports Running on http://127.0.0.1:8787 (218 actions). For diagnostic details, open Workspace → Console and look for:

[Resolve Bridge] Connected to Resolve 21.0.4.5
[Resolve Bridge] HTTP server running on http://127.0.0.1:8787
[Resolve Bridge] 218 actions available

Step 5 — Start Using It

Ask your AI assistant:

"Check if DaVinci Resolve is connected"

The assistant will call the resolve_status tool and report back. Then try:

"Create a new 1920x1080 30fps project called 'My Video' and add a timeline"

Architecture

Bridge Script (bridge/Bridge-Connection-Script.py)

A Python script that runs inside DaVinci Resolve. The supported Workflow Integration launcher starts it for the user; Workspace → Scripts → DaVinciResolveBridge remains available only as a fallback. It:

• Obtains the live resolve object from Fusion's globals
• Starts an HTTP server on 127.0.0.1:8787
• Exposes 3 endpoints:
  • GET /health — check if the bridge is alive
  • GET /actions — list all available action names
  • POST /action — execute an action by name with parameters
• Stays alive with a blocking serve_forever() call
• Kills any previous instance on the same port before starting

MCP Server (mcp_server/server.py)

A standalone MCP server that:

• Connects to the bridge over HTTP on localhost
• Auto-discovers all 218 actions from the bridge's /actions endpoint
• Exposes each action as an MCP tool with the resolve_ prefix
• Proxies tool calls to the bridge via POST /action
• Stays in sync with the bridge — no manual tool definitions needed

All 218 Tools

The MCP server auto-discovers tools from the bridge. Each tool is prefixed with resolve_ (e.g., status becomes resolve_status). Pass parameters as a params object.

Resolve Application (15 tools)

Project Manager (18 tools)

Project Settings (7 tools)

Media Storage (4 tools)

Media Pool (13 tools)

Read the full README on GitHub →