Skip to main content

Open Edit

Python License Platform MCP

AI-native video editing, driven over MCP.

Open Edit Review Studio The Review Studio, live: preview a cut, inspect renders, and scrub the edit graph.

The product timeline The timeline: clips on V1/A1/A2, edit markers, and a playhead that is ready to scrub.

Open Edit logo intro (animated GIF) *The 60-second logo intro — animated in HTML/CSS with the HyperFrames engine and rendered entirely through Open Edit's own pipeline (the agent built the project, added the overlay, and ran the final 1080p30 GPU export).*

What it is

Open Edit is an AI-native video editor that runs as a local MCP server over stdio. An external agent (Cursor, Claude Code, OpenCode, any MCP client) owns the creative loop; Open Edit owns the machinery: SHA-256 content-addressed ingest, word-level transcription (faster-whisper), an append-only IR edit graph in SQLite (WAL), and melt + ffmpeg rendering with a 640×360 proxy review artifact and a 1080p final export.

No cloud, no built-in LLM. The server is pinned to one project directory and executes 28 kinds of IR edit operations against it, each one recorded, reversible, and renderable.

Downloads

Linux / macOS — one command:

curl -fsSL https://github.com/AH64-dll/OpenEdit/releases/download/v1.3.1/install.sh | bash

Windows (PowerShell):

irm https://github.com/AH64-dll/OpenEdit/releases/download/v1.3.1/install.ps1 | iex

Source (v1.3.1): zip · tar.gz

Both installers also provision the render runtime (Node.js + the bundled HyperFrames overlay engine, plus ffmpeg/melt/Chrome checks) and print a readiness summary — see INSTALL.md → Runtime requirements.

The agent way

Open Edit is built to be operated by an agent, not a GUI. Paste the install prompt into your agent and it will clone, install, verify end to end, and report back. Then paste the configure prompt to register the MCP server in your host and confirm all six tools appear.

The normal way

Linux / macOS

git clone https://github.com/AH64-dll/OpenEdit.git
cd OpenEdit
python3 -m venv .venv
source .venv/bin/activate
pip install -U pip
pip install -e ".[mcp]"        # extras: ,serve (review UI) · ,whisper (transcription)

Windows (PowerShell)

git clone https://github.com/AH64-dll/OpenEdit.git C:\OpenEdit
cd C:\OpenEdit
python -m venv .venv
.\.venv\Scripts\Activate.ps1   # if blocked: Set-ExecutionPolicy -Scope CurrentUser RemoteSigned
python -m pip install -U pip
pip install -e ".[mcp]"

Check the entry point, create a project, and start the review studio:

.venv/bin/open-edit-mcp --help            # Windows: .\.venv\Scripts\open-edit-mcp.exe --help
mkdir -p ~/OpenEditProjects               # Windows: New-Item -ItemType Directory -Force -Path "$env:USERPROFILE\OpenEditProjects"
open_edit init ~/OpenEditProjects/my-talk
open_edit serve --review-only --port 8000 # → http://127.0.0.1:8000 (needs the ,serve extra)

Register the server in Cursor (~/.cursor/mcp.json on Linux/macOS, %USERPROFILE%\.cursor\mcp.json on Windows) and reload MCP:

// Linux / macOS
{
  "mcpServers": {
    "open-edit": {
      "command": "/ABSOLUTE/PATH/TO/OpenEdit/.venv/bin/open-edit-mcp",
      "args": ["--project", "/home/YOU/OpenEditProjects/my-talk"],
      "env": { "OPEN_EDIT_RENDER_BACKEND": "cpu" }
    }
  }
}

// Windows
{
  "mcpServers": {
    "open-edit": {
      "command": "C:\\OpenEdit\\.venv\\Scripts\\open-edit-mcp.exe",
      "args": ["--project", "C:\\Users\\YOU\\OpenEditProjects\\my-talk"],
      "env": { "OPEN_EDIT_RENDER_BACKEND": "cpu" }
    }
  }
}

What you get

Tool Role
query_project Read-only project queries
edit_project Mutations + creative generation
run_script Free-form Python IR edits (bwrap sandbox on Linux, dev on Windows)
trigger_render Enqueue proxy / final / preview-chunks renders
get_render_job Poll a durable render job by job_id
cancel_render_job Cancel queued or running jobs

Render modes: proxy (640×360 review artifact, whole-file) → final (1080p). Every render passes a deterministic 10-check QC gate before it is offered to you.

Live guide

The product is documented and illustrated in the live guide: open-edit guide — what it is, how it works, and how to install it on Linux and Windows.

Docs

  • INSTALL.md — full Linux + Windows setup, smoke checks, update/uninstall
  • docs/MCP.md — MCP tools, Cursor config, review UI, render workflow
  • skills/ — agent playbook and harness skills (also shipped in the wheel)
  • docs/REMOTION_LICENSE.md — Remotion licensing

Contributing

Bugs, ideas, and pull requests are welcome — see CONTRIBUTING.md (fork → branch → PR; CI runs tests on every pull request).

License

MIT — see LICENSE. Experimental prototype; behavior may change between releases. Motion graphics use the bundled HyperFrames engine (HTML/CSS/JS, pinned in this repo — no extra install). Remotion is legacy/migration-only. Legacy Remotion templates may require a company license — see docs/REMOTION_LICENSE.md.

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

open_edit-1.3.1.tar.gz (622.4 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

open_edit-1.3.1-py3-none-any.whl (758.9 kB view details)

Uploaded Python 3

File details

Details for the file open_edit-1.3.1.tar.gz.

File metadata

  • Download URL: open_edit-1.3.1.tar.gz
  • Upload date:
  • Size: 622.4 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/7.0.0 CPython/3.14.6

File hashes

Hashes for open_edit-1.3.1.tar.gz
Algorithm Hash digest
SHA256 1bcaa3255463fb44f590da40c5c1371ad22001eef1c4a16186131c65192eed46
MD5 319e2ee46d8e621f6d58c91312c354f0
BLAKE2b-256 f272724902cb762568d9644f49abdf71c54217a5e045209c6f5877ec77a1d0a2

See more details on using hashes here.

File details

Details for the file open_edit-1.3.1-py3-none-any.whl.

File metadata

  • Download URL: open_edit-1.3.1-py3-none-any.whl
  • Upload date:
  • Size: 758.9 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/7.0.0 CPython/3.14.6

File hashes

Hashes for open_edit-1.3.1-py3-none-any.whl
Algorithm Hash digest
SHA256 8f2fd609ff3804951f6f7c137dd8cf2be35a73ca508f02ee68fc1dbf116c0f3a
MD5 731a9e034e55df621666e76a2ddb47ac
BLAKE2b-256 b750e4388b97004671991ce3f4e1944f57a9fb4177934f3d916df646cf06e79a

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page