The Creator Funnel

← All connection guides

Advanced

Connect your AI assistant

Let your AI assistant read your website and, when your plan allows it, prepare drafts

Without it: Nothing breaks; you just work in the dashboard yourself.

AI assistants that support MCP (Model Context Protocol) can connect to your website with an access token you create. Starter includes read-only access; Growth and Pro include read + draft. They can never send email, schedule, charge money, refund, delete, publish or change settings. Those stay your click.

Several websites on one account? The billing account owner can use Account settings → AI agents for All websites in this account. This includes future websites added to that billing account, so share it only with an assistant you trust. Other billing accounts stay inaccessible, even if you own them too. This page requires two-factor login and recent authentication. Account tokens have their own active-token limit: 2 on Starter, 5 on Growth, 10 on Pro and Complimentary, separate from website tokens; account overrides can change the limit. Website-only tokens remain available for narrower access.

With an account token, tell the assistant: “First call list_websites. For every website tool call, pass the chosen website_id from that list in the arguments.” There is no remembered current website: specify it every time, including after switching sites. The client settings use the same server address and Authorization header. Listing websites uses the account safety allowance; website calls use both that account allowance and the selected website’s shared allowance. Revocation, expiry, loss of account ownership, user removal, suspension or deletion disables access; every call uses your current plan’s effective permissions.

Starter allows 2 active tokens and 30 tool calls per minute per website; Growth allows 5 and 60; Pro and Complimentary allow 10 and 120. All tokens for a website share its allowance, so adding a token does not multiply it. There is no monthly quota. A separate account safety limit defaults to 300 calls per minute across your websites. Account overrides may change these values: AI agents shows your current plan, effective access and limits. If you reach a minute limit, wait for the retry time your assistant receives.

Steps

  1. 1Create an access token

    Open AI agents with the button below. Under New token name it after the assistant, choose Read or, if available, Read + draft and an expiry, then press Create token. Copy the token: it is shown once. Active tokens have not expired or been revoked. At the cap, revoke an unused token or upgrade before adding another. A downgrade never deletes tokens or drafts; an existing draft token keeps reading but cannot draft when your current plan does not allow it.

  2. 2Add it to your assistant

    The AI agents screen shows ready-to-paste settings for Claude Code, Claude Desktop, Cursor, Hermes Agent and OpenClaw, with your server address and token filled in. Any other app that supports MCP works too: choose the Streamable HTTP connection type, paste the server address and add the header Authorization: Bearer followed by your token. Paste them into your assistant's MCP settings and restart it.

    Server address: shown in your dashboard

  3. 3Done

    Ask your assistant something like "Summarise my last 30 days". Every draft it makes shows who drafted it, so you can review it before anything goes live. Revoke a token any time on the AI agents screen.

In your dashboard: Settings → Connections → Connect your AI assistant → Set up. Provider steps checked Oct 31, 2026.