OpenClip + Cursor MCP Integration | Video Editing from Cursor - OpenClip
OpenClipOpenClip
Get started
Cursor Integration

OpenClip + Cursor MCP Integration

Cursor speaks MCP natively, which makes it a strong cockpit for video automation: wire OpenClip into mcp.json once and your editor's agent can submit recordings, poll processing, and render captioned clips as part of any workflow you script. The full config is three lines.

Cursor
AI coding agent

Prerequisites

  • An OpenClip account with an active subscription
  • Cursor with MCP support enabled

Setup Steps

1

Add OpenClip to mcp.json

Edit ~/.cursor/mcp.json for a global setup or .cursor/mcp.json inside a project, and add: {"mcpServers": {"openclip": {"url": "https://openclip.app/mcp"}}}. The URL is the same public address for everyone; no key goes in the file.

2

Reload Cursor

Restart or reload Cursor so it picks up the new server and fetches OpenClip's tool list. The tools appear under the openclip server name in Cursor's MCP panel.

3

Complete the OpenClip sign-in

The first call triggers the OAuth 2.1 flow: sign in with your OpenClip account in the browser and approve access. The grant is scoped to mcp:use and bound to your team; revoke it any time from OpenClip.

4

Drive video work from the editor

Ask Cursor's agent to submit a demo recording with submit_video, check progress with get_video_status, and render the best moments with render_clip. Because you are in an editor, it is natural to script the loop: a task that submits every new recording in a folder via create_upload and reports finished clips back.

Features

Three-line setup

One mcpServers entry in mcp.json connects the whole tool surface. Project-level config lets a repo carry its own video pipeline setup for the team.

Built for scripted loops

Cursor's agent can wrap OpenClip tools in repeatable routines: submit new recordings, poll until completed, render with the house caption preset, collect URLs.

Full tool surface

The same twenty-eight tools every client gets: submission, monitoring, clips, transcripts, caption presets, usage, reusable processing agents, and free media utilities.

No secrets in config

The mcp.json entry holds only the public URL. Authentication happens through OAuth in the browser, so nothing sensitive lands in your dotfiles or repo.

Dev-workflow native

Clip release-demo walkthroughs, turn sprint review recordings into shareable moments, or generate docs clips, without leaving the editor.

Transcripts for automation

get_transcript with millisecond windows lets scripted workflows find exact quotes or Q&A segments programmatically before rendering.

Frequently Asked Questions

Global (~/.cursor/mcp.json) if you personally use OpenClip across projects; project-level (.cursor/mcp.json) if a specific repo's workflows depend on it and teammates should get the same setup on clone. Both take the identical three-line entry.

No. The file contains only the public server URL. Auth is an OAuth sign-in in the browser on first use, and the resulting grant is revocable from OpenClip without touching the config.

Developer-adjacent video work: turning demo and walkthrough recordings into short clips, clipping conference talk VODs, or wiring a folder-watch routine that submits new recordings via create_upload and reports rendered clips back.

Yes. Both connect to https://openclip.app/mcp with their own OAuth grants, and both see the same team: videos submitted from one are visible to the other via list_videos.

Wire OpenClip into your editor

Three lines of mcp.json and a sign-in, and Cursor can turn recordings into clips as part of any workflow.

Get Started Free

Related Pages