Cloudflare Launches Traces in Open Beta with Volume-Based Pricing
Cloudflare has launched Traces in open beta, expanding automatic distributed tracing beyond Workers to cover the entire request path. The feature captures security rules, transformations, cache decisions, routing, Worker execution, and origin handling as OpenTelemetry spans within a unified timeline, requiring no manual instrumentation. This addresses a longstanding gap where traces previously showed only client and application layers, with intermediation inferred from logs and configuration.
Extending Tracing Coverage
The new spans provide timing, outcomes, and attributes for each step, enabling developers to pinpoint exactly where delays or blocks occur. For example, a trace might reveal that a security rule blocked a request and how long rule evaluation took, or show that a Transform Rule rewrote a URL before the application received it. One trace demonstrated a cache miss where 527 milliseconds of a 539-millisecond request was spent fetching from the origin.
Context propagation allows Cloudflare spans to integrate into traces initiated upstream. Traces accept a W3C traceparent header from incoming requests and can forward a new one to the origin for instrumented services. An incoming propagation policy determines whether Cloudflare accepts external trace context, mitigating noise from untrusted clients.
Volume-Based Pricing Model
The pricing model shifts from counting spans to measuring ingested data and storage. Teams already using Workers Tracing must note the change: starting December 1, 2026, billing will be based on volume rather than span count. The free plan includes 0.5 GB of daily ingestion with seven-day retention. Paid and Enterprise tiers provide 50 GB of ingestion and 10 GB-month of storage per billing cycle, with overage priced at $0.25 per GB ingested and $0.10 per GB-month stored.
Sampling rules now directly influence costs, as keeping baseline rates low reduces data volume. Trace Rules can override sampling for specific traffic, such as a customer’s hostname or requests with a debug header, while maintaining low baseline rates for general traffic. This makes cost management more predictable but requires careful rule configuration to control expenses during investigations.
Integration with Observability Tools
Spans export via OTLP to any compatible backend, with account-level destinations configurable and domain-specific routing available. It emphasizes its commitment to OpenTelemetry portability, ensuring data remains usable across observability platforms. Through the Cloudflare Observability MCP server, coding agents can query traces using SQL, compare failed versus successful traces, and link findings to repository code.
Future updates will add automatic instrumentation for DDoS rules, Access, Workflows, Queues, and Pipelines. Authenticated context propagation is planned to let trusted callers continue traces without accepting context from all clients. Other upcoming features include ad-hoc tracing for specific requests without altering baseline rates and OpenTelemetry API support in Workers for adding span attributes.
Cloudflare positions this as part of its broader observability strategy, where telemetry becomes interactive input for automation rather than passive dashboard data. The open beta is accessible via dashboard, API, or Terraform on any domain, with OTLP export configurable at the account level. While current coverage is partial by design, the roadmap indicates expanded instrumentation across HTTP request paths and Workers execution layers.
