
Shotcut MCP
io.github.matrodrigsPublished by GitHub account matrodrigsv1.6.0Updated Oct 24, 2026
Create, edit, validate, preview, and render Shotcut projects through AI clients.
Overview
Lets an AI assistant create, edit, validate, preview and render saved Shotcut video projects as editable .mlt timelines on your own machine.
- What you can do
- Create and edit timelines: tracks, clips, transitions, speed changes, titles and subtitles, with source media discovery and batched operations.
- Review edits: inspect source quality and color, preview proposed changes, and generate frames, contact sheets or short clips.
- Render and deliver: export a full project, frame range or named marker, returning the media plus the exact project used.
- Check readiness: run status and doctor checks that report tool paths, versions and failed runtime checks.
- When to use it
- Useful when you want to describe a video edit in plain language and have the assistant build or adjust a Shotcut project, review it visually, and export it, while keeping the result as an editable timeline file.
- Requirements
- Runs as a local stdio process on the user's machine (desktop only). Needs Python 3.10+, an installed Shotcut build (recommended 26.8.1 with bundled MLT 7.41.0) and an MCP client that supports local stdio servers; no extra Python packages or pip install are required. No authentication, environment variables or headers are declared.
Installation
In SourceWeft
- Open Shotcut MCP in the dashboard and add it to a workspace.
- Enable the server for the chats that should use its tools.
Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.
Other MCP clients
Follow the launch instructions in the repository.
README
Shotcut MCP
Local video editing through the Model Context Protocol.
Create, edit, validate, preview, and render saved Shotcut projects through your AI assistant. The result stays an editable .mlt timeline, with transactional writes that protect the project as you work.
Demonstration · Quick start · Documentation · Website
Demonstration
A short H.264 export created from a native Shotcut timeline edited through Shotcut MCP.
https://github.com/user-attachments/assets/c70f064f-17e7-403d-9bcf-689a9c616cdf
Quick start
You need Python 3.10+, Shotcut (recommended 26.8.1, bundled MLT 7.41.0), and an MCP client that supports local stdio servers. Shotcut includes the required media tools, codecs, and filters. No additional Python packages are required.
1. Connect your client
The installation guide includes commands for Windows, macOS, and Linux. No pip install is required to run the server. For a source checkout, the guide also shows how to check Python and Shotcut with scripts/shotcut_mcp_server.py --check before connecting MCP.
2. Check readiness
Ask your assistant:
Run shotcut_status and shotcut_doctor. Report tool paths, versions, and failed runtime checks.
Resolve failed runtime checks before editing. See troubleshooting for recovery steps.
3. Make your first edit
Provide the source folder and the destination for the saved project, then describe your edit:
You can request adjustments in ordinary language. See the editing workflow and example prompts for subtitles, quality analysis, effects, and delivery.
Features
- Edit the timeline: work with tracks, clips, transitions, speed changes, titles, and subtitles; discover source media in a folder and apply up to 500 operations in one transaction.
- Review the result: inspect source quality and color, preview proposed changes, and generate frames, contact sheets, or short clips with motion and audio before exporting.
- Render and deliver: export a full project, frame range, or named marker; receive the media and the exact editable project used for that render.
- Keep work local: use effects from your installed MLT build, with revision checks, validation, isolated backups, and atomic file replacement.
See the full capability reference for supported edit operations.
Transactional safety
Edits are checked against the inspected revision, validated with Melt, and backed up before atomically replacing the saved project. Unknown XML is preserved; ambiguous edits are rejected. Preview and render outputs also use temporary files and atomic promotion.
Each render uses an immutable project snapshot, so later timeline edits cannot change a running export. If export was not already requested, the client confirms the output, preset, range, and overwrite behavior first. An explicit export request does not need a second approval.
Read the transaction and recovery details and rendering behavior.
Limitations
- The server edits saved project files; it cannot see unsaved changes in Shotcut. Let an MCP edit finish before saving the same project in the GUI, then inspect it again.
- Available codecs, filters, fonts, and quality analyzers depend on your local installation.
- Some ambiguous transitions, speed maps, and cross-track edits are rejected. Changing FPS preserves recognized frame numbers rather than automatically retiming the edit.
- Network resources and unsafe consumer properties are denied by default. Administrators can restrict filesystem access through the runtime configuration.
See all limitations for speed-map boundaries and recovery behavior.
Documentation
MCP tools
The MCP tool catalog lists every tool with a concise purpose. Clients receive full schemas and operation guidance at runtime; most users can work entirely through natural-language prompts.
Development
See Contributing for the shared quality gate, real integration tests, development, and release procedures, Changelog for release changes, Issues for bugs and feature requests, and Security for private vulnerability reporting.
License
Released under the MIT License. This is an independent community project, not affiliated with or endorsed by Shotcut or the MLT project. Shotcut is a trademark of its respective owner; MLT is an independent open-source multimedia framework. This repository contains no Shotcut or MLT source code.
Created by matrodrigs.
Source: README.md at commit 0283c64
Tools
0Version history
2- v1.6.0LatestOct 17, 2026
- v1.5.6Sep 16, 2026
Score
50% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 days25 pts
Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.
- Auth declared10 pts
Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts