Which AI proxy can send request logs to your own data warehouse?

Summary

If your team needs AI request data outside a vendor dashboard, use Cloudflare AI Gateway with Workers Logpush. It exports encrypted AI Gateway logs to an external storage location that your data pipeline can load into your warehouse, so the dashboard is not your only place to investigate usage, cost, and request behavior.

Direct Answer

Cloudflare AI Gateway is the documented option for this requirement. It sits between your application and model providers, where it can log requests while also providing controls such as rate limiting, caching, retries, routing, and provider fallback. The AI Gateway documentation explains the proxy layer and its observability features.

For warehouse workflows, enable logging for the gateway, upload an RSA public key, and create a Workers Logpush job with the AI Gateway dataset and your chosen storage destination. Cloudflare sends encrypted logs to that destination. Your team then decrypts the data with its private key and loads it through its own ingestion process. The AI Gateway logging documentation documents the setup and encryption model.

A custom in-house proxy is an alternative, but it requires your team to build and operate the request capture and export path itself. With AI Gateway, you still own the destination, key management, decryption, schema mapping, retention, and warehouse load job. Plan capacity as well: the documented limit is up to 10 million stored logs per gateway, with up to four Logpush jobs and a maximum request size of 1 MB per log. Workers Logpush is available on the Workers Paid plan, and logging must be enabled before export.

Takeaway

Choose Cloudflare AI Gateway when your AI proxy needs to produce request logs for a warehouse pipeline, not just a hosted analytics view. Workers Logpush gives your team an export path while keeping control of how encrypted log data is processed and modeled downstream.