Davinci Resolve AI Bridge MCP

What flamexnreal/davinci-resolve-ai-bridge-mcp does

flamexnreal/davinci-resolve-ai-bridge-mcp links an MCP-compatible AI assistant with an open DaVinci Resolve project. It supports both the Free and Studio editions, although individual operations may depend on the Resolve build and Studio-only effects remain unavailable in Free.

The bridge covers project inspection, timeline navigation, media placement, clip transforms, zoom animation, selected color operations, markers, track management, and project saving. It can also create duplicate timelines for review, compare timeline structure and properties, inspect project health, and identify candidate silence intervals before explicitly accepted cuts are applied.

As a tool in the 🎥 Multimedia Process category, it fits workflows where an assistant needs structured access to an editor rather than only file-level media processing.

How it works

Start the bridge from DaVinci Resolve through Workspace → Scripts → Resolve AI Bridge → Start AI Bridge. The local worker then serves MCP requests for the active Resolve session. A session should be started once per Resolve session; selecting Start again does not create another worker, and Stop waits for an active operation to finish.

Typical first requests are resolve_status, bridge_capabilities, and timeline_overview. Status and capability checks work without an open timeline, but timeline tools require an open or newly created timeline. The bridge can use direct external scripting in Studio when local external scripting is enabled, and it falls back to the Console transport when direct attachment is unavailable.

Setup and configuration

flamexnreal/davinci-resolve-ai-bridge-mcp requires Python 3.10 or newer. Node and npm are needed for the recommended npx installer, but are not required for the base runtime. Installation creates a private environment under ~/.resolve-ai-bridge, adds Resolve menu entries and editing skills, and detects supported AI client configurations. Existing valid configurations are backed up and merged; invalid configurations are left unchanged with a diagnostic.

You can install with npx davinci-resolve-ai-bridge-mcp, use the macOS/Linux shell installer, or use the Windows PowerShell installer. A cloned checkout can be installed with the Python installer. FFmpeg is optional and can be added with the installer's --with-ffmpeg option; an FFmpeg executable already on PATH also works. After installation, reopen Resolve and refresh or restart the AI client's MCP server.

Tools and capabilities

The tool set includes:

• Inspecting connection state, Resolve version, project data, timelines, clips, tracks, markers, and settings.
• Capturing timeline frames and analyzing source audio for peaks, RMS, silence, onsets, and activity.
• Importing media, appending clips, adding images, creating timelines and compound clips, and navigating pages.
• Setting transforms, animating zoom, changing supported constant video speed, and splitting eligible clips.
• Reading and changing supported clip grades, animating saturation or color effects, and using existing Fusion blur or spotlight workflows.
• Adding markers, changing clip properties and colors, enabling or disabling clips, managing tracks, deleting explicitly requested clips, and saving the project.
• Duplicating timelines for review, comparing timelines, checking project health, and applying accepted silence cuts to isolated dialogue or aligned audio-video pairs.

The exact result can depend on Resolve's available API, decoder support, media state, and the requested workflow. bridge_capabilities reports relevant availability.

Limitations and notes

flamexnreal/davinci-resolve-ai-bridge-mcp is local and does not add a remote bridge server, public MCP tool, or manually entered token. The menu startup path has been live-tested on macOS with Resolve Free 21.0.3.7 and Python 3.14.6. Windows, Linux, and other Resolve builds require live verification; the Console fallback remains available.

Some operations are best effort. Title insertion should be checked with text_set, and clip splitting is restricted to normal-speed clips with a verified timeline checkpoint. Source-frame capture may require FFmpeg, and audio extraction can also use macOS afconvert. Closing Resolve ends the worker process.