# MCP server overview

> Connect an AI assistant to your workspace over the Model Context Protocol and run your waitlist in plain language, on the same pipeline the dashboard and API use.

Source: https://getlaunchlist.com/help/docs/mcp/overview

---

LaunchList speaks **Model Context Protocol (MCP)**, the open standard for connecting AI assistants
to live data. Connect a client once, then ask about signups, referrals and campaigns in plain
language, or have the assistant add a signup and move a contact through your funnel for you.

The MCP server runs on the **same action layer as the dashboard and the REST API**. A signup an
assistant captures is deduplicated, credited to a referrer, and run through your automations
exactly like one from your hosted page. Nothing is skipped, and nothing is a quiet database write.

## Where it answers

```text
https://mcp.getlaunchlist.com
```

Streamable HTTP transport. The same server is also reachable at `https://api.getlaunchlist.com/mcp`,
a permanent alias, so a client that only accepts an API-host URL still works.

If your client installs servers from the official MCP Registry, LaunchList is listed there as
[`com.getlaunchlist/launchlist`](https://registry.modelcontextprotocol.io/v0/servers?search=com.getlaunchlist/launchlist),
pointing at `https://mcp.getlaunchlist.com`.

## Two ways to connect

You connect with one of two credentials. Pick by the client you use.

- **Sign in (OAuth)**, recommended for chat clients like Claude and ChatGPT. Add LaunchList from
  Claude's connector directory or the ChatGPT plugin directory, or paste the endpoint into another
  client, then approve access on the consent screen. There is no key to mint or paste. An OAuth connection
  grants the full tool surface for one workspace and is revoked under **Authorized agents** in
  **Account → Security**.
- **API key**, for command-line and headless clients. Add an `Authorization: Bearer llk_…` header,
  using the same `llk_` keys the REST API uses. Keys carry scopes, so a read-only key exposes only
  the read tools. Mint one in **Embed & API → Developers → API keys**.

![Embed & API, AI assistants, with the server address for the chosen assistant and a Copy button.](https://app.getlaunchlist.com/images/docs/mcp/overview/01-ai-assistants-tab.webp "Embed & API, AI assistants shows the server address to paste into your assistant.")

Both are covered per client on [Connect your client](/help/docs/mcp/connect), and the security
model is on [Security and permissions](/help/docs/mcp/security).

## Quickstart

The fastest path is Claude Code:

```bash
claude mcp add --transport http launchlist https://mcp.getlaunchlist.com --header "Authorization: Bearer llk_live_..."
```

Then ask it something real:

```text
How many signups did I get this week?
```

## What you can do

An assistant can read your numbers and audience, and take a curated set of write actions:

- Check your numbers: signups this week, the breakdown by status and funnel stage, and where growth came from.
- Look someone up, or find everyone still waiting to verify their email.
- Look up one person's referral standing: their live position, how many they have referred, and what one more referral would move.
- Add a signup, move a contact through your funnel, or invite people to early access.
- Tag contacts, save a segment, or bulk-import a list without emailing anyone.
- Start a project, rewrite your launch page from a description, or edit the draft in progress. Putting it live is two steps: an assistant shows you the address and what goes live, and publishes only once you say yes.
- Draft a broadcast, then send it. Sending is always two steps: an assistant checks and counts first, shows you what it found, and only sends once you say yes.

The full list, with which tools read versus write, is on [Tools](/help/docs/mcp/tools).

## Next

- [Connect your client](/help/docs/mcp/connect): setup for Claude, ChatGPT, Codex, Grok, Gemini CLI, Cursor, VS Code, Windsurf, Mistral, Muse Code, OpenClaw and Hermes.
- [Tools](/help/docs/mcp/tools): every tool the assistant can call.
- [Security and permissions](/help/docs/mcp/security): scopes, consent, and how to revoke access.

## Common questions

### What is the LaunchList MCP server?

It is an endpoint that lets an AI assistant like Claude, ChatGPT, Cursor or Codex read and act on your workspace over the Model Context Protocol. It runs on the same action layer as the dashboard and REST API, so a signup captured by an assistant goes through the real pipeline.

### Do I need to write code to use it?

No. You connect a client once with a URL, then ask in plain language. Writing requests by hand is the REST API's job; the MCP server is the no-code, conversational face of it.

### Is the MCP server free?

Connecting and asking is available on every tier. Any action an assistant takes counts exactly as it would from the dashboard: a signup it captures counts toward your contact limit, and an invitation it sends is a real email.