Back to home

About OpenChatCut

OpenChatCut is a local-first, open-source, agent-native video editor: conversation-driven edits that stay editable and undoable.

Last updated: 2026-07-21

What we are building

OpenChatCut focuses on a concrete problem: let AI participate in editing, not only produce a video you cannot revise.

You describe the goal—pacing, captions, music, output size—and the agent changes a real multitrack timeline. Every step is visible. You can refine it or undo it. The project remains yours.

Principles

Local-first

Video media is large and often private. Projects and media stay on your machine by default (~/.openchatcut, with IndexedDB as a browser cache). Online models, media search, and remote storage are optional add-ons. A project does not require a cloud account just to exist.

Open and auditable

The product is open source. You can read the code, build it yourself, and audit how agents change the timeline and where keys and data live. License details are in the GitHub repository.

Model-agnostic

The built-in agent supports multiple providers and OpenAI-compatible APIs. External agents such as Codex and Claude Code can drive the same project through MCP. Switching models should not force you to rebuild the editing workflow.

Edits must stay editable

Generative tools often treat one-shot output as the end state. OpenChatCut turns every agent action into real timeline changes: tracks, captions, transitions, effects, keyframes, and history remain. People and agents share the same editor capabilities.

Who it is for

  • Creators and teams who frequently cut interviews, talking-head videos, podcasts, and product demos
  • People already using agents like Claude or Codex who want them on a real timeline
  • Professionals who care about local data and undoable workflows
  • Developers exploring MCP or agent-native editing

Community and source

Related reading