Video Editing MCP Server: What It Is and How to Connect - OpenClip
MCP Glossary

Video Editing MCP Server

A video editing MCP server turns an AI assistant into a video producer. Connect once with OAuth at https://openclip.app/mcp and your assistant can submit source videos, track processing, review detected moments, and render captioned vertical clips, all through typed tool calls instead of a browser.

Definition

A video editing MCP server is a Model Context Protocol server that exposes video production actions as typed tools an AI assistant can call. Where a general MCP server might expose files or databases, a video editing server publishes the verbs of a clipping pipeline. OpenClip's production server is a reference example: submission tools (submit_video for URLs, create_upload and complete_upload for local files), monitoring tools (get_video_status, list_videos, get_render_status), output tools (list_clips for scored viral moments, render_clip for captioned rendering, get_transcript for time-coded text, list_caption_presets for styling options, list_supported_providers for accepted sources), account tools (get_account, get_usage, where credits are minutes of processing), and configuration tools (list_agents, create_agent, update_agent, describe_agent_settings, plus logo tools for watermarking). The workflow is asynchronous: the assistant submits a job, polls status every 10 to 15 seconds, presents the detected moments for approval, and renders the ones you choose. Connection is a single public URL, https://openclip.app/mcp, secured with OAuth 2.1 and dynamic client registration, so any MCP client from Claude to Cursor connects with a sign-in rather than an API key; header-only agents can use a personal access token at https://openclip.app/mcp/key instead.

Related Terms

Features

Submission tools

submit_video takes a source URL (YouTube, Vimeo, direct links and more via list_supported_providers); create_upload and complete_upload handle local files through presigned uploads.

Monitoring tools

get_video_status and get_render_status return structured job states, so the assistant reports real progress and catches failures like download_failed or pending_credits with the right fix.

Clip and caption tools

list_clips returns the detected viral moments with virality scores; render_clip produces a captioned vertical clip using any style from list_caption_presets.

Transcript access

get_transcript returns time-coded segments at sentence or word level, with millisecond windowing, so assistants can find quotes and Q&A moments before clipping.

Reusable processing presets

create_agent and update_agent save caption preset and composition settings, so every episode of a show gets identical treatment without re-briefing.

OAuth 2.1 security model

One public server URL, per-user OAuth sign-in with a scoped mcp:use token, and video-derived text marked as untrusted content so assistants never execute instructions hidden in a video.

Frequently Asked Questions

Add the server URL to your client and sign in. For OpenClip that is https://openclip.app/mcp: in Claude it is Settings, Connectors, Add custom connector; in Claude Code it is one command, claude mcp add --transport http openclip https://openclip.app/mcp; in Cursor it is an mcpServers entry in mcp.json. OAuth handles the rest.

No. OpenClip's server uses OAuth 2.1 with dynamic client registration, so you approve access from inside the assistant. Agents that can only send headers can mint a personal access token and use the https://openclip.app/mcp/key endpoint.

Processing is asynchronous. OpenClip downloads the source, transcribes it, detects viral moments, and tracks speakers while your assistant polls get_video_status. On completion, list_clips returns scored moments and render_clip produces the captioned clips you approve.

Processing uses credits counted in minutes, included with OpenClip subscriptions: Starter includes 150 processing minutes per month and Pro includes 300. An active subscription is required; a video that fails immediately after submission usually means there is none.

Yes. The server URL is the same for every client, and each user signs in with their own account. You can connect Claude, ChatGPT, Cursor, and Perplexity side by side; jobs and credits belong to your team either way.

They solve different problems. An API is for code you write; an MCP server is for assistants that operate tools conversationally, with runtime discovery and per-user OAuth. Many teams use MCP for day-to-day operation and an API for deep custom integrations.

Connect your assistant to OpenClip

One URL, one sign-in, twenty-eight video tools. Turn long videos into captioned clips from the assistant you already use.

Related Pages