Streaming server logs — Cloudflare, Vercel, CloudFront, and Signals

AI crawler analytics (LLM Site Visibility) needs to see your real traffic. Maya accepts it from whichever layer you already have — your CDN, your host, or your own servers. All sources feed the same pipeline; pick one.

Cloudflare

In LLM Site Visibility → Config, paste a Cloudflare API Token and your Zone ID, then hit Test Connection. Maya pulls AI-bot traffic for the zone from then on. Use a token scoped to analytics read access for the zone, nothing broader.

Vercel

Connect a Log Drain from your Vercel project. Vercel pushes request logs to Maya as they happen; AI-bot requests are filtered and aggregated on arrival.

AWS CloudFront

CloudFront standard logs (v2) delivered through Amazon Data Firehose push to Maya's logs endpoint with your streaming credentials. One-time setup in AWS; near-real-time after that. The step-by-step CloudFront guide is in the developer docs.

Signals / direct streaming

For everything else there are streaming credentials (project + source + write key) against logs.withmaya.ai:

• Point a log shipper — Vector, Fluent Bit, or similar — at the endpoint from Nginx, Apache, IIS, or any custom stack.
• Or install the lightweight JS collector snippet where server logs aren't available.
• Run the built-in test event to verify the pipe before trusting it.

Per-stack guides (Nginx, Apache, IIS, custom BFF endpoint) with exact filter configs are in the developer docs.

Which should you pick?

You run	Use
Cloudflare in front of the site	Cloudflare token — 2 minutes, no infra
The site on Vercel	Log Drain — native push
CloudFront CDN	Standard logs v2 + Firehose
Your own servers / other CDN	Log shipper to logs.withmaya.ai

One source is enough; adding a second for the same traffic double-counts. Choose the layer closest to the edge (the one that sees bot requests your origin might never receive).

Privacy

The pipeline is built for bot analytics: guides include source-side filters so you ship crawler/agent traffic, not your users' browsing. What Maya does and doesn't collect is documented in the security docs.