MCP server

Let the agent that wrote your app fill in its pages

Connect Claude Code, Cursor or any MCP client to Lanpage. It reads your app's own code to answer the questions your documents depend on, drafts the pages and the landing, and tells you what the stores still require. Nothing goes live until you say so.

Connect it once

Add the Lanpage server to your client with the option that matches it.
The first time, the client opens your browser: sign in and approve what it may do. You never paste a key.
Ask your agent to set up the app. You can disconnect it at any time from AI agents, in your account.

Claude Code
```
claude mcp add --transport http lanpage https://lanpage.co/api/mcp
```
Run it in a terminal, then /mcp inside Claude Code to sign in.
Claude (web and desktop)
```
https://lanpage.co/api/mcp
```
Settings → Connectors → Add custom connector, and paste this address.
Cursor, Windsurf and other clients with an mcp.json
```
{
  "mcpServers": {
    "lanpage": {
      "url": "https://lanpage.co/api/mcp"
    }
  }
}
```
Add this to the client's MCP configuration file.

VS Code

code --add-mcp '{"name":"lanpage","type":"http","url":"https://lanpage.co/api/mcp"}'

Run it in a terminal, or add the server from the MCP view.

Clients without remote servers
```
npx mcp-remote https://lanpage.co/api/mcp
```
A local bridge for clients that only start local MCP servers.

What an agent can do

Answer the app facts (accounts, analytics, ads, payments, AI) from the repository.
Write drafts of the Privacy Policy, Terms, support and account deletion pages, and custom pages.
Edit the landing: blocks, theme, FAQ, store links, logo and screenshots.
Read the store checklist and say what still blocks a submission.
On a paid plan, publish, unpublish or change the address, after you confirm it.

What stops it from acting on its own

Everything it writes is a draft. What is live only changes when you publish.
Publishing, deleting and changing an address wait for your own confirmation, in the client or on a one-time link. The model cannot confirm for you.
It only reaches the account you signed in with, under the same limits as the dashboard.
It never sees the email addresses your landing collected, your billing, or how to close your account.
Your account lists every connected client and every change an agent made, and disconnecting one stops it at once.

What you approve when you connect

The approval screen shows which client is connecting, to which account and which address it returns to, and the permissions it asks for:

lanpage:read: Read your apps, their pages, checklist and traffic
lanpage:write: Edit drafts: app details, landing, FAQ, images and documents
lanpage:publish: Publish, unpublish, delete and change addresses, each after you confirm

Approving a permission skips no limit: your plan and the ownership of each app are checked on every call, as in the dashboard.

How you confirm a change

Publishing, unpublishing, showing or hiding a document, deleting and changing an address do not happen until you answer. Declining or not answering changes nothing.
If your client can ask you questions, the confirmation appears in the conversation.
If it cannot, the agent hands you a one-time link. Open it with the account you connected: it lasts 30 minutes and there you confirm or decline. Meanwhile, AI agents lists it as “Sent you a confirmation link”.
Exactly what you were asked is applied: if the agent changed the draft while you decided, nothing is published and it has to ask you again.

If something does not work

Claude Code does not show the Lanpage tools or says it needs authentication: type /mcp, pick lanpage and sign in again.
The agent says the connection lacks a permission: you disconnected it or that permission was not approved. Authenticate again from the client and approve what it asks for.
Publishing answers that it needs a paid plan: on Free the agent reads and drafts, and you publish from the dashboard.
The confirmation link expired or was already used: ask the agent to repeat the action and it will give you a new one.
The agent gets “too many requests”: each account allows 120 calls per minute. Wait a minute and carry on.
Your client only starts local servers: use the npx mcp-remote bridge above.

On every plan

Reading and drafting work on every plan, including Free. Publishing from an agent is part of Pro; on Free you publish from the dashboard, which stays free.

Compare every plan side by side