MCP server

Choppity runs a remote MCP server, so an AI assistant can use your Choppity workspace directly: make clips from a YouTube link, a video file link or a file on your computer, check on them, render them and post them to your connected socials.

Server URL: https://api2.choppity.com/mcp (Streamable HTTP)

Auth: either sign in (OAuth 2.1 — Claude, ChatGPT, Claude Code and any client that supports MCP authorization), or an API key from Settings → API sent as Authorization: Key <your key>. Either way, only workspace owners on a paid plan can connect.

Connect

Claude (claude.ai and Claude Desktop)

1. Open Settings → Connectors and choose Add custom connector.
2. Name it Choppity and paste https://api2.choppity.com/mcp.
3. Click Connect, sign in to Choppity if asked, and click Approve.

ChatGPT

1. Open Settings → Apps and create an app. If you don’t see the option, turn on Developer mode under Advanced.
2. Paste https://api2.choppity.com/mcp and choose OAuth.
3. Sign in to Choppity if asked, and click Approve.

Claude Code

claude mcp add --transport http choppity https://api2.choppity.com/mcp

Then start claude, type /mcp, pick choppity and choose Authenticate. To use an API key instead (for scripts and CI), add --header "Authorization: Key $CHOPPITY_KEY".

Cursor

In ~/.cursor/mcp.json:

{
  "mcpServers": {
    "choppity": {
      "url": "https://api2.choppity.com/mcp",
      "headers": { "Authorization": "Key YOUR_KEY" }
    }
  }
}

VS Code

Run MCP: Open User Configuration from the command palette, add this, and use Copilot Chat in Agent mode:

{
  "servers": {
    "choppity": {
      "type": "http",
      "url": "https://api2.choppity.com/mcp",
      "headers": { "Authorization": "Key YOUR_KEY" }
    }
  }
}

Codex

In ~/.codex/config.toml, then restart Codex:

[mcp_servers.choppity]
url = "https://api2.choppity.com/mcp"
http_headers = { "Authorization" = "Key YOUR_KEY" }

Any other client that speaks Streamable HTTP works the same way — with a sign-in if it supports MCP authorization, or with a key header if it doesn’t.

Disconnect

Apps you signed in with are listed under Settings → MCP → Connected apps. Disconnect cuts an app off within a minute. Access also stops if the workspace moves to the free plan or you stop being its owner.

Tools

Tool	What it does	Changes anything
whoami	Shows the workspace the key acts for	no
start_clip_job	Makes clips from a YouTube link, a video file link or an upload_id. Uses your monthly analysis allowance.	yes
get_clip_job	A job’s status and its clips: a link to open each clip in Choppity, and a low-resolution preview that plays in the browser	no
start_render	Full-quality render of one clip. Uses your rendering allowance.	yes
get_render	A render’s status and download link	no
create_upload	Starts uploading a file from your computer (up to 10 GB)	yes
refresh_upload_urls	New upload links for parts whose links expired	yes
complete_upload	Finishes an upload	yes
list_socials	Your connected social accounts	no
list_clip_posts	A clip’s post drafts and posts	no
get_post	One post	no
update_post	Edits a draft’s caption, title or settings	yes
publish_post	Posts publicly, now or at a scheduled time	yes
cancel_post	Cancels a scheduled post or discards a draft	yes

Tools are labelled read-only or destructive, so assistants that respect those labels ask before posting, editing or cancelling.

Uploading a file from your computer

The server can’t read your disk, but an assistant with a terminal can:

1. create_upload with the file’s exact size and type returns an upload_id and a short bash script.
2. The assistant saves the script and runs it: bash upload.sh ./recording.mp4 > parts.json. It uploads the file in parts (one part in memory at a time) and prints the parts list.
3. complete_upload with that parts list.
4. start_clip_job with the upload_id. Each upload feeds one clip job; to clip the same file again, upload it again.

Upload links last 4 hours. For a slow connection, refresh_upload_urls re-issues the ones that ran out.

Limits

The MCP server shares the rate limits of the REST API, per workspace: 60 requests a minute, and for get_clip_job, get_render, get_post and list_clip_posts the tighter status limits. A limited call comes back as a tool error that says how many seconds to wait. While a job or render is running, its status includes retry_after_seconds.