Local-first editing Open source

Edit at the speed of thought.
Stay native to Shotcut.

Ask your MCP client to trim clips, add titles, relink media, or render a delivery. Shotcut MCP applies structured edits to the saved .mlt file, validates them locally, and keeps the result editable in Shotcut.

Install for your client Download latest release

Shotcut MCP demonstration Native timeline edit to H.264 export

24 FPS 1080 P H.264 EXPORT

01 — Intent“Trim the pause and export the marked range.”
02 — TransactionNative .mlt operations
03 — ResultValidated. Backed up. Editable.

Works with saved Shotcut projects MLT 7.41 Python 3.10+ Windows, macOS, and Linux

Editing workflow

From intent to an editable project.

Shotcut MCP operates on the project format itself. It does not imitate mouse clicks or hide the changes behind a hosted service.

01
Inspect the starting point
Discover source media or inspect the saved project, using focused pages for large timelines.
inspect_project → revision
02
Plan when useful, then apply
Dry-run the operations, review the XML diff and Melt validation result, then submit up to 500 related operations as one atomic edit.
plan → edit × 1…500
03
Validate, recheck, and commit
Validate the temporary project with Melt, recheck the on-disk revision, create an isolated backup, then atomically replace the original.
validate → recheck → save
04
Review, render, or continue in Shotcut
Review the result visually, start a durable export, or open the native project for hands-on adjustments. Reinspect after saving in Shotcut.
preview | render | open

Project safety

Edits fail safely or complete cleanly.

Revision checks reject stale candidates. A per-project lock serializes MCP commits. Melt validation, isolated backups, and atomic replacement keep the last valid project on disk until the new one is ready.

Read the safety model

1
Guard the baseRevision check · project lock
2
Validate the candidateIn-memory edit · Melt · recheck
3
Promote safelyIsolated backup · atomic replace

Tool coverage

What the server can actually do.

Use focused MCP tools for discovery, editing, review, rendering, and recovery. Inspect locally installed MLT services when a filter or transition needs lower-level control.

TIMELINE

Build and revise

Build tracks with clips, gaps, titles, markers, and subtitles
Refine timing with trim, split, move, roll, slip, slide, and speed
Apply transitions, ordered filters, and native MLT keyframes

REVIEW

Inspect, review, and recover

Inspect structure, resources, timing, and revisions before editing
Review source quality, exact frames, contact sheets, and clips with audio
Relink missing media or restore a validated isolated backup

DELIVERY

Validate and deliver

Validate MLT services, color workflows, and available encoders
Render full timelines or exact ranges with durable status and history
Deliver chapter text and locally supported video or audio presets

See all 28 MCP tools

Installation

Choose the shortest path to a working server.

Choose your MCP client, then copy the shortest supported setup. Every option runs the same dependency-free Python server locally over stdio.

Default stack: Shotcut 26.8.1 · MLT 7.41.0 (7.41.x family) · Python 3.10+

Windows integration also covers Shotcut 26.6.25 / MLT 7.40.0. Other version pairs receive an untested-version warning.

Project boundary: Saved .mlt files; unsaved GUI state stays in Shotcut.

Before registration, run the entry point with --check. See the installation guide if Windows opens the Microsoft Store instead of Python.

Claude Code · Plugin marketplace

Install from the conversation.

Add the marketplace, install the plugin, then reload it without restarting Claude Code.

/plugin marketplace add matrodrigs/shotcut-mcp
/plugin install shotcut-mcp@matrodrigs
/reload-plugins

Codex · Local stdio server

Clone once, then register.

Choose your platform. The command points Codex at the same dependency-free Python server.

WindowsPowerShell · Python

git clone https://github.com/matrodrigs/shotcut-mcp.git
cd shotcut-mcp
$pythonExe = py -3 -c "import sys; print(sys.executable)"
& $pythonExe "C:\path\to\shotcut-mcp\scripts\shotcut_mcp_server.py" --check
codex mcp add shotcut -- $pythonExe "C:\path\to\shotcut-mcp\scripts\shotcut_mcp_server.py"

macOS or LinuxTerminal · Python 3

git clone https://github.com/matrodrigs/shotcut-mcp.git
cd shotcut-mcp
python3 /absolute/path/to/shotcut-mcp/scripts/shotcut_mcp_server.py --check
codex mcp add shotcut -- python3 /absolute/path/to/shotcut-mcp/scripts/shotcut_mcp_server.py

Other MCP clients

Use the package or standard stdio.

Start with the packaged release if your client supports MCPB. Otherwise clone the repository and configure the local server directly.

View release

Nameshotcut

CommandAbsolute Python 3.10+ executable path

Argument/absolute/path/to/shotcut-mcp/scripts/shotcut_mcp_server.py

Common questions

Local files. Editable projects.

Does Shotcut MCP control the Shotcut interface?

No. It edits saved MLT project files directly and uses the local Shotcut, Melt, FFmpeg, and FFprobe executables for validation, previews, and rendering.

Are my videos uploaded anywhere?

The server processes media locally and sends tool results, metadata, and preview images to your MCP client. That client may use a cloud model or upload data under its own policies. Network resources embedded in projects are denied by default.

Can I keep editing the result in Shotcut?

Yes. The output remains a native, editable .mlt project. Let the MCP finish its transaction before reopening or saving the same project in Shotcut.

Is it affiliated with Shotcut or MLT?

No. Shotcut MCP is an independent MIT-licensed community project and is not endorsed by Shotcut or the MLT project.

Install locally

Choose your client. Keep the project native.

Start with a saved Shotcut timeline or a folder of source media. Review the result before exporting.

Choose an install path Download the release

Edit at the speed of thought.Stay native to Shotcut.