MCP server

The welaunch MCP server

welaunch finds where people who need what you sell are already looking, on ChatGPT, Google, Reddit and X, and writes the posts, replies and pages that bring them to you. As an MCP server, it does that from any agent that speaks MCP: Claude Code, Cursor, Windsurf, the Claude app, or one you built. Ask it to run a play, read the five replies it wrote, and send the ones you like to your browser.

Connect

On every plan, Free included. Reading is free; the tools that run agents spend credits exactly as the app does, and each connection has a daily credit cap.

ChatGPT and Claude (OAuth, no key)

Add a custom connector with the URL below. In Claude: Settings, Connectors, Add custom connector. In ChatGPT: turn on developer mode in Settings, then create a connector. You sign in to welaunch, pick the product, and choose Allow. The server follows the MCP authorization spec: OAuth 2.1 with PKCE, dynamic client registration, and the protected resource and authorization server metadata documents.

https://api.welaunch.sh/mcp

Claude Code, Cursor, Windsurf (API key)

Make a key in Settings, MCP server. Then:

Claude Code

claude mcp add --transport http welaunch https://api.welaunch.sh/mcp --header "Authorization: Bearer wl_YOUR_KEY"

Cursor, Windsurf, the Claude app: add to mcp.json

{
  "mcpServers": {
    "welaunch": {
      "url": "https://api.welaunch.sh/mcp",
      "headers": {
        "Authorization": "Bearer wl_YOUR_KEY"
      }
    }
  }
}

What you can ask

• "What are my top plays this week?"
• "Run the Reddit agent and show me the replies."
• "Rewrite the third one, less salesy."
• "Send the first two to my browser."
• "What is still waiting on me?"
• "Score competitor.com and tell me what they get wrong."
• "How is my visibility on ChatGPT compared to last week?"

The tools

Two rules

Nothing is posted from the MCP. post_target queues the draft for the welaunch Browser Agent, the Chrome extension in your own browser, which opens the page, fills the reply in, and waits for you to press submit. An assistant with a tool that posts to Reddit on its own is how accounts get banned; yours never has one.

Every key has a daily credit ceiling, 50 by default, set per key in Settings. A step that finds and writes costs 8 credits; an agent in a loop stops at the cap, not at your balance.

Webhooks

Add an https endpoint in Settings, Webhooks and pick events. Each event is a POST with a JSON body { type, timestamp, data }, where data always carries workspaceId. Answer with any 2xx. A delivery is tried three times: at once, a minute later, and ten minutes after that. The log in Settings shows what each attempt got back.

Signed the Standard Webhooks way, so any of its libraries verifies them. Three headers: webhook-id (the same on every retry, so you can drop duplicates), webhook-timestamp (Unix seconds) and webhook-signature, which is v1, and a base64 HMAC-SHA256 of id.timestamp.body keyed with your secret. By hand, in Node:

import crypto from "node:crypto";

// secret: the whsec_... value shown once when you added the endpoint.
// body: the raw request body, as a string, before any JSON parsing.
function verify(secret, headers, body) {
  const id = headers["webhook-id"];
  const ts = headers["webhook-timestamp"];
  if (Math.abs(Date.now() / 1000 - Number(ts)) > 300) return false; // replay
  const key = Buffer.from(secret.replace(/^whsec_/, ""), "base64");
  const want = crypto.createHmac("sha256", key).update(`${id}.${ts}.${body}`).digest("base64");
  return headers["webhook-signature"].split(" ").some((part) => {
    const [v, sig] = part.split(",");
    return v === "v1" && sig.length === want.length &&
      crypto.timingSafeEqual(Buffer.from(sig), Buffer.from(want));
  });
}

The skill

Agents that load skills can take /skills/welaunch/SKILL.md: when to reach for welaunch, how to read a play, and how to talk about drafts (as drafts, until the founder presses submit).