renezander030/capcut-cli

What it solves

CapCut and JianYing projects are typically edited manually in a GUI. capcut-cli provides a way to create and edit these projects programmatically via a command-line interface, allowing for automation and integration with LLM agents without needing an official API or a running server.

How it works

The tool operates directly on the local draft store of CapCut/JianYing, which uses JSON files to store project data. It loads these JSON files, validates them against a version-aware schema, applies the requested edits, and writes the changes back to the disk atomically. This ensures that the resulting projects remain fully editable within the CapCut/JianYing applications.

Who it’s for

• AI Agent Builders: Developers creating agents that can perform video editing tasks (e.g., automated captioning, cutting, and translation).
• Content Creators: Editors who want to automate repetitive tasks like silence-aware cuts and subtitle imports.
• Developers: Those looking to programmatically generate video project files from JSON specifications.

Highlights

• Agent-Ready: Includes a "skill" installer for agents like Claude Code and Cursor, and an experimental Wasm-based "look, don't touch" mode for safer AI review.
• Zero Dependencies: Works as a standalone binary with no need for an MCP server, HTTP daemon, or cloud uploads.
• Cuts and Captions: Features specialized commands for silence detection, scene-cut detection, and multi-language translation of drafts.
• Full Editability: Unlike flattened exports, it produces real project files that can be opened and adjusted manually in the CapCut app.
• Cross-Platform: Supports both CapCut and JianYing namespaces on macOS and Windows.