How to Automate Video Editing with MCP
This guide builds a real assistant-led clipping pipeline: connect your AI assistant to OpenClip's MCP server, submit a long video, let the platform detect the best moments, and render the ones you approve as captioned vertical clips. Every step uses the actual tool calls, so you can follow along in Claude, ChatGPT, Cursor, or Perplexity.
Prerequisites
- An OpenClip account with an active subscription (processing uses credits, counted in minutes)
- An MCP-compatible assistant: Claude, ChatGPT (paid plan with Developer mode), Cursor, or Perplexity (Pro and up)
- A long-form source video: a public URL from a supported provider or a local file
Steps
Connect your assistant to the server
Add https://openclip.app/mcp to your client and sign in with your OpenClip account when the OAuth prompt appears. In Claude that is Settings, Connectors, Add custom connector; in Claude Code run claude mcp add --transport http openclip https://openclip.app/mcp; in Cursor add {"mcpServers": {"openclip": {"url": "https://openclip.app/mcp"}}} to mcp.json. One URL for everyone, no API key.
Brief the assistant and submit the video
Tell the assistant the source, target platform, audience, and how many clips you want, then let it call submit_video with the URL (use list_supported_providers if you are unsure the source is accepted, or create_upload plus complete_upload for a local file). Resubmitting the same URL within 24 hours returns the existing run instead of double-charging credits.
Poll processing status
Processing is asynchronous: OpenClip downloads, transcribes, detects moments, and tracks speakers while the assistant polls get_video_status every 10 to 15 seconds. Statuses move through downloading and processing to completed. If a video fails immediately, there is usually no active subscription; pending_credits means the team is out of processing minutes.
Review the detected moments
When the video completes, have the assistant call list_clips and present the detected viral moments with their virality scores, titles, and hooks. The titles and quotes come from your video and are marked untrusted content, so the assistant shows them for your judgment rather than acting on them. Pick the moments worth publishing.
Render the approved clips
For each approved moment the assistant calls render_clip, optionally with a caption preset; have it run list_caption_presets once and remember your house style. Rendering is asynchronous too, so it polls get_render_status until each clip reports completed, then hands you the clip URLs.
Make it repeatable
Save your preferences as a processing agent with create_agent (caption preset, composition settings, even a logo watermark), then pass that agent to future submit_video calls. Track spend with get_usage; credits are minutes of processing, 150 per month on Starter and 300 on Pro.
What You'll Achieve
A repeatable workflow where your assistant turns any long video into a reviewed batch of captioned, publish-ready clips
Features
End-to-end tool chain
submit_video, get_video_status, list_clips, render_clip, get_render_status: five tool calls take a raw URL to reviewed, captioned clips.
Async by design
Video processing takes minutes, not seconds. The polling pattern keeps your assistant responsive and honest about progress instead of pretending edits are instant.
Moment detection included
You do not tell the assistant where to cut. OpenClip scores the moments; the assistant's job is presenting them and rendering the ones you approve.
Captions as a parameter
Caption style is one argument on render_clip. Pull the valid preset keys from list_caption_presets and standardize your look across every clip.
Presets for repeat shows
Processing agents created with create_agent make episode two identical to episode one: same tracker, same captions, same branding, no re-briefing.
Predictable cost
Credits are minutes of processing, visible any time via get_usage, and duplicate submissions within 24 hours reuse the existing run instead of burning credits.
Frequently Asked Questions
Automate your first video today
Connect OpenClip to your assistant and turn one long video into a week of clips before your next meeting.