API keys and the MCP server

Let your own tools — scripts or AI assistants like Claude — read and edit your site, safely scoped.

Everything you can do in the editor, your own tools can do over an API — including AI assistants that speak the Model Context Protocol (MCP), like Claude.

Mint a key

Go to Dashboard → Account → API keys and create a key. Three dials keep it safe:

  • Scopes — what the key may do (read content, edit content, publish, analytics…). Give a key only what it needs.
  • Site pinning — optionally lock the key to a single site.
  • Expiry — optionally give it an end date.

The key (it starts with rl_) is shown once — store it in a password manager. A key can never do more than your role on a site allows: an editor's key can't publish, no matter its scopes. Revoke keys any time; what keys do is recorded in the site's audit log.

Connect an AI assistant (MCP)

The platform has a built-in MCP server. In your assistant's connector settings add:

  • URL: your builder's address followed by /api/mcp
  • Authentication: Bearer — paste your rl_ key

Then just talk: "list my sites", "add a FAQ section to the services page", "draft a post about our spring fundraiser", "how many orders came in this week?". The assistant sees tools for pages and sections, blog posts, media, products and orders, collections, redirects and analytics — always limited by your key's scopes and your role.

Plain REST

Prefer scripts? The same key works as a Bearer token against the REST endpoints your dashboard uses. Ask your platform operator for the endpoint reference if you're building something custom.

Webhooks are the other direction — the site calling your systems when something happens (a form arrives, an order is paid). An admin configures those per site under Integrations.

Was this article helpful?