Roundtable × Cloudflare Pages — Bot‑Gate Drop‑In (Beta)

Adds Roundtable's auto‑block gate to any existing Cloudflare Pages project. The block‑list lives in Cloudflare KV, so a hot lookup adds only ~1 ms of latency.

Setup should take 15 minutes max.

1. Embed the Roundtable tracker

<script
  src="https://cdn.roundtable.ai/v1/rt.js"
  data-site-key="YOUR_SITE_KEY"
  data-auto-block="true"   <!-- **required** so the gate can call /rt-block -->
  data-user-id="OPTIONAL_USER_ID">
</script>

Get YOUR_SITE_KEY at https://accounts.roundtable.ai/account/keys. The full tracker docs are available at https://docs.roundtable.ai.

2. Install & initialize

# Add dev dependency
npm i -D roundtable-cloudflare-bot-gate

# Scaffold functions and generate token
npx roundtable-cloudflare-bot-gate init

init does two things:

1. Adds functions/_middleware.js and functions/rt-block.js.
2. Generates a 64‑char RT_WEBHOOK_TOKEN and prints it.

3. Store the secret in Cloudflare

echo <RT_WEBHOOK_TOKEN> | wrangler secret put RT_WEBHOOK_TOKEN

Use the token printed by init.

4. Create KV namespaces & add binding

Create the namespaces and copy the generated IDs:

wrangler kv namespace create RT_BLOCKED            # copy the production ID
wrangler kv namespace create RT_BLOCKED --preview  # copy the preview ID

Then add the binding to your Wrangler config:

For wrangler.jsonc:

{
  "kv_namespaces": [
    {
      "binding": "RT_BLOCKED",
      "id": "abcd1234...",        // paste production ID here
      "preview_id": "wxyz5678..."  // paste preview ID here
    }
  ]
}

For wrangler.toml:

[[kv_namespaces]]
binding = "RT_BLOCKED"
id = "abcd1234..."        # paste production ID here
preview_id = "wxyz5678..."  # paste preview ID here

5. Email your token + webhook URL (beta step)

Send the token printed in step 2 and your webhook endpoint to [email protected] so we can enable blocking for your site.

Endpoint:  https://<your‑domain>/rt-block
Token:     <RT_WEBHOOK_TOKEN>

6. Deploy

wrangler pages deploy

Roundtable will now write sid:<id> into RT_BLOCKED when it detects fraud; the middleware returns 403 on the very next request.

Need help?

Email [email protected] or open an issue.