MCP.so
ログイン

Notch MCP Server認証済み注目

@Notch

基本情報

トランスポート

Streamable HTTP

認証方式

OAuth

カテゴリ

メディアとデザイン

作者

Notch

投稿者

Ajitesh Choudhary

Notch MCP Server について

Create finished video ads from chat. Give Notch a brief or product URL — it loads your brand context, writes the script, and produces the finished ad: avatar, scenes, voiceover, captions. Built for performance marketers, DTC founders, and agencies. Works with Claude, Perplexity, Claude Code, and any MCP client.

接続情報

https://app.usenotch.ai/mcp

セットアップ

claude mcp add notch --transport http https://app.usenotch.ai/mcp

Notch MCP Server のツール

9

List the Notch workspaces the user belongs to. New ad sessions are created and billed in the active workspace (marked isActive). Use select_workspace to switch.

Switch the active Notch workspace — where new ad sessions are created and billed. Pass a workspaceId from list_workspaces.

List recent ad sessions in the active workspace, most recently active first. Any of these can be resumed with send_message or get_session_updates.

Create a new Notch ad session from a plain-language brief and start the agent working. Returns the sessionId to poll with get_session_updates, plus the workspace and credit balance to relay to the user. Fails fast (before any work) if the workspace can't afford the minimum billable render. If the account has several workspaces and hasn't picked one for this connection, no ad is created and the result asks which workspace to use — relay that question, then call again with workspaceId, or useActiveWorkspace when they explicitly confirm the current one.

Long-poll for what's new in an ad session: status, progress lines, milestone media, agent questions, and final deliverables. The server tracks what it already sent you — each call returns only new items, and always carries the current status plus any question the agent is waiting on. Holds server-side and returns as soon as there is something to report — call it again immediately while status is "working".

Continue an ad session's conversation: answer the agent's questions, request edits, or submit an intake form. After sending, poll get_session_updates for the agent's response.

Stop the agent's in-flight work on an ad session — call this when the user asks to stop, cancel, or abandon the ad. The session itself is kept: a later send_message resumes it, and work the agent already finished stays done.

Open an upload box so the user can put one of their local files (a product photo, a reference image or video) into the ad. Use this whenever the user has a file on their device and this client can show interactive UI: they drop the file in the box, it uploads straight to Notch, and its URL comes back to you to pass in attachments. Never ask the user to attach a local file to the chat instead — chat attachments can't be read; the box is the way in. Wait for the user to finish the upload before using the returned URL.

Get a file into an ad when you hold the bytes yourself, returning a URL to pass in attachments. In a code or terminal environment, call it for a presigned URL and PUT the bytes there. For a very small file you may instead pass its base64 contents as fileData to store it in one step. When the file is on the user's device in a chat client, use upload_file instead so they upload it through a box.

概要

What is Notch?

Notch is an agentic ad-creation platform. Connect it to Claude (or any MCP client) and you can go from a one-line brief to a finished video ad without leaving the conversation — Notch loads your brand's context, writes the script, builds the storyboard, casts the avatar, and produces the final video.

What you can do

  • Brief in plain language — "Make a 30-second UGC-style ad for our summer drop, targeting Gen Z"
  • Use your brand automatically — colors, tone, product details, and approved hooks are pre-loaded from your Notch brand kit; no re-briefing every session
  • Full production pipeline — script → storyboard → avatar → scenes → voiceover → edited video
  • Direct revisions in chat — "punchier hook", "shorter", "swap the middle scene"

How to connect (Claude)

  1. In Claude, go to Settings → Connectors → Add custom connector
  2. Name it Notch and paste: https://app.usenotch.ai/mcp
  3. Click Connect and sign in with your Notch account

Works the same way in Perplexity (custom remote connector), Claude Code, and any MCP client that supports remote servers.

Config for JSON-based clients:

{
  "mcpServers": {
    "notch": { "type": "http", "url": "https://app.usenotch.ai/mcp" }
  }
}

Requirements

A Notch account — sign up at usenotch.ai.

Links

よくある質問

1

Notch のリモート MCP サーバーとは何ですか?

Notch のリモート MCP サーバーは、https://app.usenotch.ai/mcp にホストされた Model Context Protocol エンドポイントです。AI アシスタントはローカルに何もインストール・実行することなく接続できます。

2

Notch の MCP サーバーにはどう接続しますか?

エンドポイント https://app.usenotch.ai/mcp を Claude Code、Cursor、VS Code など任意の MCP 対応クライアントに追加してください。このページのセットアップ手順で各クライアントを一度に設定できます。

3

Notch の MCP サーバーは認証が必要ですか?

はい。Notch は OAuth を使用します。初回接続時に MCP クライアントがブラウザを開いてサインイン・認可を行い、以降はその認証情報を再利用します。

4

Notch の MCP サーバーはどのトランスポートを使用しますか?

Notch は Streamable HTTP エンドポイントを公開しており、これはリモート MCP サーバーで広く使われ、主要な MCP クライアントがサポートするトランスポートです。

コメント