Cloudflare engineering blog

Cloudflare announces eight observability updates that unify logs, traces, analytics, alerts, dashboards, SQL querying, and telemetry export across its platform. The release adds end-to-end tracing, agent-accessible observability APIs, custom alerting, OpenTelemetry support, and usage-based pricing changes.

Cloudflare introduces a Web Search API for AI Gateway, integrating Ceramic.ai, Exa, and Linkup to provide live web context for model inference and agents. The post explains REST API and Workers integrations, observability, billing, BYOK, zero-data-retention indicators, and crawler transparency requirements.

Cloudflare reports progress toward making formerly Enterprise-only capabilities available across its plans, including Logpush, RBAC, multi-account organizations, Terraform management, and resource tagging. The update also details expanded platform limits, internal account-management practices, and planned availability for upcoming features.

Cloudflare highlights how civil society organizations are using serverless infrastructure, Workers AI, and AI Gateway to automate editorial workflows, mental health support, human-rights research, and policy tracking. The examples emphasize cost control, security, privacy, data minimization, and human review when deploying AI for sensitive work.

Cloudflare introduces Traces, an open-beta distributed tracing service that follows requests through security rules, transformations, caching, routing, Workers, and origins. It supports sampling rules, W3C trace context propagation, OpenTelemetry/OTLP export, dashboard investigation, and agent-assisted debugging.

Cloudflare introduces an Account Abuse Protection dashboard for investigating suspicious login and signup activity. It uses stateful behavioral context, privacy-preserving Hashed User IDs, network and device signals, and investigative filters to help teams identify compromised accounts and apply targeted responses.

Cloudflare explains how Protected Quick Tunnels add accountless email authentication through the new --allowed-mail flag. The design combines Cloudflare Access for identity verification with a stateless broker and local authorization in cloudflared, keeping guest lists on the developer’s machine while avoiding per-request policy lookups.

Cloudflare reports that it is the fastest provider across 74% of the world’s 1,000 largest networks, up from 60% in April 2026. The post explains its trimean connection-time methodology and how privacy-preserving measurements from Challenge Pages expand real-user performance data and improve ranking confidence.

Cloudflare explains Streamline, an open-source architecture for long-running custom video pipelines built with Workers, Containers, and Durable Objects. The post covers session lifecycle management, media ingestion and output over RTMPS, HLS, and WebSockets, pipeline operations, preview delivery, and security controls.

Cloudflare introduces a managed OHTTP Gateway that separates relay-based client identity from encrypted request contents, preserving OHTTP’s double-blind privacy model. The post explains its edge deployment, HPKE key management, relay authentication, chunked request support, abuse protections, and when to choose a gateway over a relay.

Cloudflare introduces Workers KV Instant, a Quicksilver-powered mode that delivers sub-2 ms p99 reads and near-real-time replication across more than 300 edge locations. The post explains its API compatibility, pricing, workload constraints, operational limits, and trade-offs versus classic Workers KV.

Cloudflare announces general availability for AI Search, adding native image embeddings, OCR for scanned PDFs, support for files up to 10 MiB, and compatibility with any chat model. The post explains its multimodal retrieval pipeline, hybrid search flow, and usage-based pricing.

Cloudflare introduces a fully managed option for Cloudflare OS, an agent workspace that connects organizational context, data, and systems. The update also adds GitHub repository workflows, Google Workspace integration, and exports to formats including Excel, CSV, PDF, Markdown, and HTML.

Cloudflare introduces an open-beta Artifacts platform for building Git-like workflows designed for large numbers of AI agents. The post details programmable repositories, Workers bindings, event subscriptions, deployment previews, data-jurisdiction controls, metrics, and a competition to build the next agent-oriented Git platform.

Cloudflare Workers adds opt-in Web Crypto support for ML-KEM and ML-DSA post-quantum primitives, including key encapsulation, signatures, public-key derivation, JWK handling, and capability detection. The post explains how these APIs support JWT, HPKE, and OHTTP integrations while outlining compatibility, implementation, and migration limitations.

Cloudflare introduces Basin, a generally available serverless analytics platform built on Apache Iceberg and R2 Object Storage. The post explains its ingestion pipelines, managed catalog maintenance, distributed SQL engine, data portability, scalability improvements, and zero-egress pricing model.

Cloudflare introduces Clef and Clef-flash, open-source decision models designed for fast, typed classification in agentic workflows. The post details their schema-bound architecture, benchmark and latency results, Qwen-based training, calibrated reinforcement learning, and a platform for customer-specific fine-tuning.

Cloudflare introduces K2, a serverless event-streaming service built on R2 object storage for durable, ordered logs, long-term retention, and independent consumer scaling. The post explains its partitioned-segment architecture, batching trade-offs, latency characteristics, subscription model, and differences from Queues and Pipelines.

Cloudflare outlines its approach to sovereign AI through open multilingual models, including EuroLLM and Apertus, and model-agnostic security tooling. The post also describes an orchestration harness for coordinating models to find, verify, and prioritize vulnerabilities, alongside workshops for public-sector cyber teams.

Cloudflare describes how AI agents are reshaping web traffic, separating training crawlers, search bots, and task-oriented agents. It presents Web Bot Auth, configurable crawler controls, Pay Per Use, Monetization Gateway, x402 payments, and agent-optimized formats as infrastructure for visibility, access control, and web monetization.