Cloudflare exposes proxy traffic as OpenTelemetry spans
Cloudflare has launched Traces in open beta, automatically converting proxy operations into OpenTelemetry spans to eliminate visibility gaps between client requests and origin servers.

Cloudflare has opened a public beta for Cloudflare Traces, bringing automatic OpenTelemetry span generation to its entire proxy pipeline. Without requiring extra code instrumentation, security evaluation, transform rules, cache decisions, Worker execution, and origin handling are formatted into a single request timeline. In one scenario cited by Cloudflare, a cache miss revealed that 527ms of a 539ms total request duration was spent waiting for an origin server response.
The service supports W3C traceparent headers for context propagation across upstream callers and origin services, governed by configurable incoming propagation policies. Tracing relies on Cloudflare's rules engine, allowing developers to set baseline sampling rates, such as 1%, and write Trace Rules to target specific hostnames, paths, IP addresses, or debug headers. Export happens via OTLP to external observability backends, while AI coding agents can query traces using SQL through the Cloudflare Observability MCP server.
Alongside the beta release, Cloudflare updated its telemetry pricing structure starting December 1, 2026, replacing a previously planned per-span billing model with volume-based metrics. The free tier offers 0.5GB of daily ingestion with seven-day retention. Paid and Enterprise plans include 50GB of ingestion and 10GB-month of storage per billing cycle, with additional usage billed at $0.25 per GB ingested and $0.10 per GB-month stored.
Though current coverage excludes DDoS rules, Access, Workflows, Queues, and Pipelines, Cloudflare plans to extend automatic instrumentation to these features alongside authenticated context propagation and an OpenTelemetry API for Workers. Developers can access Traces today through the Cloudflare dashboard, API, or Terraform across all domains.
This is our own summary of reporting by InfoQ AI



