resolve-mcp-kit
Provides tools to inspect and control DaVinci Resolve projects, media pools, timelines, pages, settings, markers, titles, and rendering via Resolve's scripting API.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@resolve-mcp-kitlist my projects and show the current timeline"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
resolve-mcp-kit
A small, standalone MCP (Model Context Protocol) server that lets an AI agent see and drive DaVinci Resolve on your own computer. It works with any MCP client: opencode, Claude Code, Claude Desktop, and others. Beyond the Python packages in requirements.txt it needs nothing else.
It also ships an opencode skill (and a Claude Code copy) that teaches the agent how to build a timeline from clips, make cuts, add titles, and render.
What it can do
Always available (read-only, plus changing the visible page):
Tool | Purpose |
| Is Resolve reachable; product, version, page, project, timeline |
| Projects in the current database folder |
| Timelines and their frame range, tracks, fps |
| Clips in a bin: name, duration, fps, resolution |
| Switch to media, cut, edit, fusion, color, fairlight, deliver |
| Read project settings such as frame rate and resolution |
Editing tools, only when RESOLVE_MCP_ALLOW_WRITE=1 (nothing here deletes
anything):
Tool | Purpose |
| Import files into the media pool, optionally into a bin |
| New empty timeline, made current |
| Append clips in order, optionally trimmed (cuts) |
| Insert a Text+ title and set its text |
| Timeline marker |
| Save |
| Queue, start and watch a render |
Provenance: the read-only tools are the ones from the author's original server. The editing tools are new in this kit and have only been tested against a fake Resolve object (see Testing).
Related MCP server: mcp-da-vinci
Requirements
DaVinci Resolve with external scripting. Historically the paid Studio edition is required; Blackmagic has restricted scripting in the free edition in many versions. Check your version, and expect "Could not attach" in the free one.
In Resolve: Preferences > System > General > External scripting using: Local.
Resolve must be running with a project open in its window. The scripting API only attaches to a Resolve on the same computer, so this server runs on the same machine as Resolve.
Python 3.10 or newer (the scripting module needs a Python that matches what Resolve supports; 3.10 to 3.12 are the safe choices).
pip install -r requirements.txt(the start scripts do this for you). Notemcpis pinned below version 2 on purpose: this code uses the 1.xFastMCPAPI.
Where Resolve's scripting module lives
The server finds it automatically from these defaults. Override with environment variables if your install is elsewhere.
OS |
|
|
Windows |
|
|
macOS |
|
|
Linux |
|
|
Windows and Linux paths are as documented by Blackmagic and as used by the original author on Windows. The macOS and Linux defaults come from Blackmagic's README and have not been tried by the author.
Quick start
Windows:
Open DaVinci Resolve and a project.
Double-click
start-resolve-mcp.cmd(first run creates.venvand installs dependencies). It listens onhttp://127.0.0.1:8930/sse.
macOS / Linux: ./start-resolve-mcp.sh (needs chmod +x once).
Check it: with the server running, ask your client to call resolve_status. If
you see connected: false, the reason field says what to fix.
Configuration
All by environment variable (edit the top of the start script, or set them yourself).
Variable | Default | Meaning |
|
|
|
|
| Bind address for sse/http |
|
| Port for sse/http |
| unset | If set, requests need |
|
| Register editing and render tools |
| per OS | Resolve scripting paths |
Endpoints: sse serves /sse; streamable-http serves /mcp.
Security warning
This server can open, change and render your projects and write files wherever
Resolve can. Keep it on 127.0.0.1 whenever you can. If the client is on
another machine, bind to a LAN or VPN address only (for example a Tailscale
address), set RESOLVE_MCP_TOKEN to a long random string, and never expose
the port to the internet or forward it on your router. The token is sent in
cleartext over plain HTTP, so it protects against casual access, not against
someone who can watch your network; use a VPN for anything beyond your own LAN.
Connecting a client
Two styles. stdio: the client launches the server itself (simplest; the client must run on the Resolve machine). Network (sse/streamable-http): you start the server with the start script and the client connects by URL, which also works from another machine.
opencode
Config goes in ~/.config/opencode/opencode.json (global) or opencode.json in
a project folder. Merge the mcp block from one of the examples:
examples/opencode.local.json- opencode launches the server over stdio. Edit the two paths to your install of this folder.examples/opencode.remote.json- connect to the running server by URL. If you setRESOLVE_MCP_TOKEN, export the same value as an environment variable before starting opencode; the example reads it with{env:RESOLVE_MCP_TOKEN}. If you did not set a token, delete theheadersblock. For another machine, change the URL to its address.
Then run opencode mcp list to confirm resolve connects.
Install the skill (the folder name must match the skill name, per opencode):
Global: copy
skills/davinci-resolve-editingto~/.config/opencode/skills/davinci-resolve-editing(Windows:%USERPROFILE%\.config\opencode\skills\davinci-resolve-editing).Per project: copy it to
<project>/.opencode/skills/davinci-resolve-editing.
opencode also reads .claude/skills/ and .agents/skills/. Start a new
opencode session and ask it to "edit video in Resolve"; the skill is loaded on
demand.
Claude Code
claude mcp add resolve --transport sse http://127.0.0.1:8930/sseWith a token, add --header "Authorization: Bearer YOUR_TOKEN". Or use stdio:
claude mcp add resolve -e RESOLVE_MCP_TRANSPORT=stdio -e RESOLVE_MCP_ALLOW_WRITE=1 -- C:\path\to\resolve-mcp-kit\.venv\Scripts\python.exe C:\path\to\resolve-mcp-kit\resolve_mcp_server.pySkill: copy skills/davinci-resolve-editing to ~/.claude/skills/ (global) or
<project>/.claude/skills/. A copy is already at
.claude/skills/davinci-resolve-editing in this folder, so opening Claude Code
here picks it up.
Claude Desktop
Merge examples/claude_desktop_config.json into your Claude Desktop config
(Settings > Developer > Edit Config), fix the paths, and restart the app. It
uses stdio.
Other clients
Point any MCP client at http://HOST:8930/sse (SSE), or use
RESOLVE_MCP_TRANSPORT=streamable-http and http://HOST:8930/mcp.
Running at login (optional, Windows)
Resolve's API only works in your logged-in desktop session, so start the server
with a Task Scheduler task "At log on of " running
start-resolve-mcp.cmd, not as a service.
Testing
pip install -r requirements.txt pytest
python -m pytest testsThe tests replace Resolve's scripting module with a fake. They prove the server
starts, lists its tools, the read/write split works, and that the edit and
render tools call the API with the expected arguments. They cannot prove that a
real Resolve accepts those calls. Things to expect to need a real-install
shakedown: title text setting (add_title), append_clips trimming, and codec
keys for start_render.
License
MIT, see LICENSE. DaVinci Resolve is a trademark of Blackmagic Design; this project is not affiliated with it.
This server cannot be deployed
Maintenance
Related MCP Connectors
AI video editor for agents and humans: timeline, captions, color, audio and generation as MCP tools.
Generate, extend and edit AI images and videos, then assemble clips on a timeline and render.
Agentic AI video editor: edit a real timeline, generate clips, music and voice, then render.
Create and manage cinematic AI video renders through the Future Video Studio Agent API.
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceExposes 20 tools for project, media, render, and timeline control in DaVinci Resolve 18+ via stdio MCP, enabling AI agents to operate the software through natural language.40 PyPI53MIT
- AlicenseNot gradedqualityDmaintenanceExposes the DaVinci Resolve scripting API to LLMs, enabling inspection, media import, timeline editing, markers, and render queue management through natural language.MIT
- AlicenseAqualityAmaintenanceEnables AI assistants to control DaVinci Resolve (Free and Studio) through 215 MCP tools covering the full Resolve scripting API, including projects, timelines, media, Fusion, color, and rendering.178 PyPI1MIT
- AlicenseNot gradedqualityBmaintenanceEnables AI assistants to control DaVinci Resolve Studio through the official Scripting API, covering editing, media pool management, rendering, grading, Fusion, Fairlight, and project lifecycle tasks.6,981 npmMIT