Connect the MCP server

Your MCP token connects your AI tool to one website. Apps (Claude, ChatGPT) take the token in the URL. Code tools (Claude Code, Cursor…) take it in an Authorization: Bearer header.

Get your token

  1. Open the dashboard at /app/tokens.
  2. Copy your MCP token.

Keep the token secret. It can write pages. If it leaks, rotate it in the dashboard. The old token stops working right away.

One token = one website. For 3 websites, add 3 connectors.

Two ways to send the token

Your toolURLToken
Claude (web, desktop, mobile), ChatGPThttps://mcp.flocon-ai.com/<token>In the URL
Claude Code, Cursor, VS Code, Windsurf, Codex, Gemini CLI, your own agenthttps://mcp.flocon-ai.comAuthorization: Bearer <token> header

Why two ways? Claude and ChatGPT connectors cannot send custom headers. So the token goes in the URL.

The URL with the token is a password. Do not share it, paste it in a chat or commit it. If it leaks, rotate the token.

Claude (web, desktop and mobile apps)

Open the connector settings

In Claude (claude.ai or the desktop app), go to Settings then Connectors.

Add a custom connector

Click Add custom connector. Name it flocon-ai. Use this URL (replace <token> with your MCP token):

https://mcp.flocon-ai.com/<token>

Leave the OAuth fields (Client ID, Client Secret) empty.

Check it works

Start a new chat. Ask Claude: "Call get_website from Flocon AI."

A connector you add on the web also works in the desktop and mobile apps.

ChatGPT

Turn on developer mode

In ChatGPT, open Settings, then Apps & Connectors (or Connectors), then Advanced settings. Turn on Developer mode. Your plan must allow it.

Create the connector

Click Create. Name it flocon-ai. Use this URL:

https://mcp.flocon-ai.com/<token>

Authentication: No authentication (the token is already in the URL).

Check it works

In a new chat, turn on developer mode and the flocon-ai connector. Ask: "Call get_website from Flocon AI."

Claude Code

Run this in your terminal:

claude mcp add --transport http flocon-ai https://mcp.flocon-ai.com --header "Authorization: Bearer <token>"

Add --scope user to use it in every project.

Or share it with your team in .mcp.json at the project root. Each person sets FLOCON_MCP_TOKEN in their shell, so the token stays out of git:

.mcp.json
{
"mcpServers": {
  "flocon-ai": {
    "type": "http",
    "url": "https://mcp.flocon-ai.com",
    "headers": {
      "Authorization": "Bearer ${FLOCON_MCP_TOKEN}"
    }
  }
}
}

Then ask Claude Code: "Call get_website from Flocon AI."

Cursor

Add this to .cursor/mcp.json (or your global ~/.cursor/mcp.json):

.cursor/mcp.json
{
"mcpServers": {
  "flocon-ai": {
    "url": "https://mcp.flocon-ai.com",
    "headers": {
      "Authorization": "Bearer <token>"
    }
  }
}
}

Then ask Cursor: "Call get_website from Flocon AI."

VS Code (GitHub Copilot)

Add this to .vscode/mcp.json. VS Code asks for the token the first time and stores it safely. Note: the top key is servers, not mcpServers.

.vscode/mcp.json
{
"inputs": [
  {
    "type": "promptString",
    "id": "flocon-token",
    "description": "Flocon AI MCP token",
    "password": true
  }
],
"servers": {
  "flocon-ai": {
    "type": "http",
    "url": "https://mcp.flocon-ai.com",
    "headers": {
      "Authorization": "Bearer ${input:flocon-token}"
    }
  }
}
}

Then, in Copilot Chat (Agent mode), ask: "Call get_website from Flocon AI."

Windsurf

Add this to ~/.codeium/windsurf/mcp_config.json (Cascade → MCPs → View raw config). Windsurf uses serverUrl, not url.

mcp_config.json
{
"mcpServers": {
  "flocon-ai": {
    "serverUrl": "https://mcp.flocon-ai.com",
    "headers": {
      "Authorization": "Bearer <token>"
    }
  }
}
}

Refresh the MCP list in Cascade. Then ask: "Call get_website from Flocon AI."

Codex (OpenAI)

Codex reads the token from an environment variable. It never writes it to disk.

export FLOCON_MCP_TOKEN="<token>"
codex mcp add flocon-ai --url https://mcp.flocon-ai.com --bearer-token-env-var FLOCON_MCP_TOKEN

This writes the server to ~/.codex/config.toml. Put the export line in your shell profile (~/.zshrc or ~/.bashrc) so it is set every time. Check with /mcp inside Codex.

Gemini CLI

Add this to ~/.gemini/settings.json (or .gemini/settings.json in your project):

settings.json
{
"mcpServers": {
  "flocon-ai": {
    "httpUrl": "https://mcp.flocon-ai.com",
    "headers": {
      "Authorization": "Bearer <token>"
    }
  }
}
}

Restart Gemini CLI. Check with /mcp.

Your own agent (Claude Agent SDK)

Pass the server in mcpServers. Keep the token in an environment variable:

import { query } from "@anthropic-ai/claude-agent-sdk";

for await (const message of query({
prompt: "Call get_website from Flocon AI.",
options: {
  mcpServers: {
    "flocon-ai": {
      type: "http",
      url: "https://mcp.flocon-ai.com",
      headers: { Authorization: `Bearer ${process.env.FLOCON_MCP_TOKEN}` },
    },
  },
},
})) {
console.log(message);
}

Another tool?

Any MCP client that speaks Streamable HTTP works:

  • It can send headers → URL https://mcp.flocon-ai.com + header Authorization: Bearer <token>.
  • It cannot send headers → URL https://mcp.flocon-ai.com/<token>, no auth.

What you should see

get_website returns your website name, languages, plan and page counts. If it does, you are connected.

Troubleshooting

  • 401 error: the token is wrong or was rotated. Copy the new token from /app/tokens.
  • 401 error with "This is a website token (FLOCON_TOKEN). Your AI tool needs the MCP token": you pasted the website token. Copy the MCP token from /app/tokens.
  • Claude or ChatGPT cannot connect, or asks you to sign in: you used https://mcp.flocon-ai.com without the token. Use https://mcp.flocon-ai.com/<token>.
  • The header is ignored: check the word Bearer and the space before the token: Bearer <token>.
  • Wrong website: each token belongs to one website. Use the token of the website you want.
  • Do not use the website token. It is for your website (FLOCON_TOKEN). The MCP refuses it.

Next: Writing good pages.

Updated