Drive Wabcut from scripts and AI agents
Every edit is a typed command you can run from the command line, from an AI agent over MCP, or from a JavaScript extension.
Every edit in Wabcut is a typed command: the same commands the command palette and the Assistant use. You can run them from outside the app too.
Run commands from the command line#
wabcut-cli comes with Wabcut: on Windows it is in the bin folder next to wabcut.exe, in the installed app and
in the portable ZIP.
| Command | What it does |
|---|---|
wabcut-cli cmd <project> --json '<request>' | Runs commands on a saved project. |
wabcut-cli tools | Prints the commands as tool definitions for an LLM. |
wabcut-cli mcp --project <file.wabcut> | Serves the project to Claude, Codex and other agents over MCP. |
wabcut-cli export-timeline <project> <out> --format fcp7-xml|fcpxml|otio|edl | Exports a timeline for another editor. |
Let an AI agent edit with you#
wabcut-cli mcp serves one project to agents such as Claude Desktop, Claude Code, Codex and Cursor over the Model
Context Protocol.
Add your own commands#
Extensions add commands to Wabcut in JavaScript. Each runs in its own sandboxed process and gets only the capabilities you grant it.
Go further#
The developer documentation has the details:
- Edit commands: requests, results and a reference page for every command;
- MCP server: setup for each agent, and the safety model;
- Extensions: the manifest, permissions and the API.
See also
- Use the command palette and the AssistantRun any edit by name from the palette, or describe it in your own words and let the Assistant propose it.
- Use AI on your computer or with your own keyEvery AI feature is optional. Local models send nothing anywhere; cloud models use your own key, ask first and show the price.