Supercut Platform API and MCP docs:

https://api.supercut.ai/api/platform/v1/docs

Supercut MCP works in Claude Desktop, Claude.ai, and Claude Cowork, as well as Claude Code, Cursor, VS Code, Codex, and other MCP clients. Supercut is also listed in Claude's connector directory, so in Claude you can connect directly from Settings > Connectors without entering a server URL manually. For a step-by-step walkthrough, see Supercut Claude Connector.

MCP (Model Context Protocol) lets AI tools like Claude, Cursor, and Codex connect directly to Supercut, so they can pull your recordings, transcripts, comments, and more into a conversation. See MCP Tools for the full list of what an agent can do once it's connected.

1. Connect with OAuth (recommended)

Point your MCP client at the Supercut server and sign in to Supercut in your browser when it asks. You choose the workspace to grant, and the client stores the credential — there's nothing to create or paste.

Supercut uses remote HTTP MCP. Server URL:

https://mcp.supercut.ai/mcp

If your client doesn't support browser sign-in, use a personal or workspace API token instead — see 4. Fallback: API tokens below.

2. Set up your client

Claude (Desktop, Claude.ai, Cowork)

Install Supercut from the Claude Connector Directory, then sign in and pick a workspace. No URL or token needed. See Supercut Claude Connector for the full walkthrough.

Claude Code

claude mcp add \
  --scope user \
  --transport http \
  supercut https://mcp.supercut.ai/mcp

This only writes the config — it doesn't sign you in. In Claude Code, run /mcp. If Supercut says Needs authentication, choose Authenticate and finish in the browser. If it already says Connected, you're done. This is the flow we've verified end to end.

Cursor

{
  "mcpServers": {
    "supercut": {
      "type": "http",
      "url": "https://mcp.supercut.ai/mcp"
    }
  }
}

Add this to .cursor/mcp.json (project) or ~/.cursor/mcp.json (global), or add the server via Cursor Settings > Tools & MCPs, then enable Supercut. If Cursor asks you to connect, finish the browser flow: sign in, choose a workspace, and approve. Cursor doesn't always open the browser automatically on first use — if OAuth never starts, use the token fallback below.

Codex

[mcp_servers.supercut]
url = "https://mcp.supercut.ai/mcp"

Then run codex mcp login supercut and finish in the browser.

VS Code

{
  "servers": {
    "supercut": {
      "type": "http",
      "url": "https://mcp.supercut.ai/mcp"
    }
  }
}

Add this to .vscode/mcp.json, or use Command Palette > MCP: Add Server > HTTP. Start the server, accept workspace trust, then sign in in the browser if prompted.

Full, up-to-date setup steps for each client also live at the Platform API docs and at supercut.ai/agents/getting-started/cursor (swap cursor for claude-code, claude-desktop, codex, or vscode). Using Opencode? Its config lives in the Platform API docs under MCP client setup.

3. Test the connection

Open your AI tool and ask it to use Supercut. The same prompts work whether you connected with OAuth or a token:

Use the Supercut MCP to list recent recordings I have access to and extract any bug reports to work on

Or, if you have a specific Supercut share ID:

Fetch the transcript and comments for this Supercut: <share id>

If the connection is working, your AI tool should be able to retrieve Supercut data through MCP.

For everything you can do once connected, see MCP Tools.

4. Fallback: API tokens

If your client doesn't support browser sign-in, create an API token and send it as a bearer token instead:

Authorization: Bearer <token>

In Cursor, that means adding a headers field to the same server entry in mcp.json instead of a bare URL.

Personal API token

Acts as you, in one workspace: recordings you own, recordings shared with you, and recordings you've watched. This is the preferred fallback when a client needs a pasted credential.

To create one, open your workspace settings and go to Personal API tokens (/settings/{workspace}/personal-api-tokens):

Click Create token.
Give it a name, such as: Claude MCP
Choose an expiration date.
Click Create token and copy it — you'll only see the full token once.

Personal access is no longer scope-based — the create-token dialog only asks for a name and an expiry.

Workspace API token

Workspace tokens can't list or search recordings, and can't use add-playlist-recording or list-form-responses — they can only reach recordings in that workspace that you already have a share ID for, using tools that don't need a user identity: get-recording, get-transcript, get-frame, list-comments, list-reactions, and get-recording-analytics. They can also read playlists shared workspace-wide, including listing them: list-playlists, get-playlist, and list-playlist-recordings.

To create one, open your workspace settings and go to Workspace API tokens (/settings/{workspace}/api-tokens) — same steps as above: name, expiration, create, copy.

Personal tokens and OAuth both require the personal API token feature on your workspace plan/seat. If it's missing, Supercut will say the plan or seat doesn't include creating personal tokens — worth checking before troubleshooting a failed sign-in.

5. Manage or revoke access

Open your workspace settings and go to Personal API tokens. Alongside your tokens you'll find an OAuth credentials section listing every client you've signed in with — that's where you disconnect a client, like Claude Code or Cursor, without deleting a pasted token.

That's it!

Your AI tool can now read your Supercuts directly, and add a recording to a playlist. See MCP Tools for everything it can do once connected: pull transcripts, summarise recordings, surface comments and reactions, check analytics or form responses, and manage playlists.

Model Context Protocol (MCP)