Cloudflare

Cloudflare AI News & Updates51 Updates

The latest AI news and updates of Cloudflare — Network and security company offering AI inference, edge computing, and developer platform services. Covering Cloudflare's latest product updates, launches, and company news from the past 90 days.

CloudflareCloudflareOct 21

Cloudflare Enables WebMCP Shopping Tools for 100,000 Commerce.com Stores

Cloudflare deployed Google's WebMCP standard across nearly 100,000 Commerce.com storefronts, providing AI agents with direct tools for product search and cart management. The integration uses edge-side script injection to bypass traditional HTML scraping, while allowing individual merchants to opt out via custom configuration rules. This update enables agents to interact with stores without requiring site-owner code changes.

Read more
CloudflareCloudflareOct 21

Cloudflare Partners With Mastercard on Agent Pay for Machines

Cloudflare partners with Mastercard on Agent Pay for Machines, a system enabling AI agents to pay for digital services like models and compute. The platform enforces defined budgets and permissions, ensuring agents transact securely at machine speed. Early access provides buyer- and seller-side APIs for building and testing agent-to-agent commerce experiences.

Read more
CloudflareCloudflareOct 20

Cloudflare API MCP Server Adds Step-Up Authorization Support

Cloudflare’s API MCP server now supports step-up authorization, allowing agents to request additional permissions only when a specific task requires them. When a tool call triggers an insufficient_scope challenge, the MCP client prompts for approval while retaining previously granted scopes. This flow is available now in all MCP clients that support the step-up specification.

Read more
CloudflareCloudflareOct 19

Cloudflare Cuts Internal AI Code Reviewer Costs by 68 Percent

Cloudflare reduced its internal AI code reviewer costs by 68 percent after identifying that inefficient prompt caching and model usage caused expenses to double. By optimizing cache breakpoints and adjusting model effort, the system now averages $0.77 per run. These changes also improved median review speeds, cutting the time from over 4 minutes to 1 minute 32 seconds.

Read more
CloudflareCloudflareOct 19

Cloudflare Launches Clairic, a System One Extraction Model for Typed Values

Cloudflare launched Clairic, a System One extraction model that pulls, computes, estimates, and forecasts typed values from unstructured text and images. The model operates in a single parallel pass without generating tokens, charging only for input. Clairic is available as a preview on Workers AI, providing grounded, well-formed outputs with confidence scores and probability metrics for each field.

Read more
CloudflareCloudflareOct 15

Cloudflare and OpenAI Share Findings from AI Search Discovery Pilot

Cloudflare and OpenAI released findings from a joint research pilot using real-time network signals to improve AI search discovery. The project, which utilized 1% of Cloudflare’s network, helped OpenAI’s crawler index hundreds of millions of new pages. The signals filter out dead or duplicate content and flag potentially harmful pages, improving search accuracy without increasing infrastructure load.

Read more
CloudflareCloudflareOct 10

Cloudflare Launches Multimodal Clef-omni, Lowers Clef-flash Pricing, and Boosts Speed

Cloudflare launched Clef-omni, a multimodal decision model natively processing audio, video, images, and text in a single pipeline. The update also reduces Clef-flash pricing to $0.038 per million input tokens and improves Clef inference speeds by up to 2.0x. These changes expand the Clef family’s capabilities for agentic workflows.

Read more
CloudflareCloudflareOct 8

Cloudflare Launches Agentic Security Harness for Managed Defense Alerts

Cloudflare launched an evidence-grounded agentic security harness for Managed Defense to reduce alert fatigue. The system enforces scope boundaries in code before inference and uses Clef for initial triage. For deeper analysis, it coordinates four specialist AI agents powered by OpenAI’s GPT-5.6 Cyber and Anthropic’s Mythos to deliver consolidated, evidence-backed security advisories.

Read more
CloudflareCloudflareOct 2

Cloudflare AI Gateway Adds Native Web Search API Integration

Cloudflare AI Gateway now supports native web search API integration in partnership with Ceramic.ai, Exa, and Linkup. This update injects real-time web context into model inference calls via AI Gateway, REST APIs, or Workers bindings. The integration supports Bring-Your-Own-Key and adheres to Cloudflare’s bot-crawling standards for transparency and content control.

Read more
CloudflareCloudflareOct 1

Cloudflare Releases Clef and Clef-flash Open-Source Decision Models

Cloudflare launched Clef and Clef-flash, its first in-house trained models, as open-source decision models hosted on Workers AI. These models provide high-speed classification and structured outputs for agentic workflows, featuring a 64K context window and vision encoder. Cloudflare also debuted a reinforcement learning platform to help customers fine-tune these models for specific use cases.

Read more
CloudflareCloudflareOct 1

Cloudflare Adds European Open Models and AI Security Tools

Cloudflare is expanding its sovereign AI offerings with two European open-source models, EuroLLM and Apertus, now available on Workers AI. The company also open-sourced a model-agnostic security harness for vulnerability hunting and launched a workshop program for government cyber agencies, starting in Singapore this October. These tools provide nations with alternatives to frontier model dependency.

Read more
CloudflareCloudflareOct 1

Cloudflare Opens Waitlist for Fully Managed Cloudflare OS Agent Workspaces

Cloudflare is opening a waitlist for fully managed Cloudflare OS deployments, allowing organizations to launch agent workspaces in a few clicks. The managed service handles configuration and maintenance, while new updates enable agents to mount Git repositories, research Google Workspace threads, and export work into formats like Excel, CSV, and PDF.

Read more
CloudflareCloudflareOct 1

Cloudflare AI Search Reaches General Availability with Multimodal Capabilities

Cloudflare AI Search is now generally available, adding native image-pixel embedding for visual retrieval and OCR for scanned PDFs. The service now accepts files up to 10 MiB and integrates with any chat model. Billing for the platform begins November 1, 2026, with a free monthly tier available on all Workers plans.

Read more
CloudflareCloudflareSep 30

Cloudflare Launches Pay Per Use Beta for AI Content Compensation

Cloudflare launched its Pay Per Use beta, a system for publishers to earn revenue when AI products use their content. AI companies define and price specific uses, while publishers choose which offers to accept. Cloudflare handles enrollment, usage reporting via API, and monthly payouts, providing a direct commercial relationship between content owners and AI buyers.

Read more
CloudflareCloudflareSep 30

Cloudflare Releases Auto Router for AI Gateway in Public Beta

Cloudflare released Auto Router in public beta, a new feature for AI Gateway that automatically routes requests to the most capable model for a given task. The system uses an edge-deployed classifier to balance output quality against token costs, delivering up to 30% savings compared to using frontier models alone. The feature is free during the beta period.

Read more
CloudflareCloudflareSep 30

Cloudflare Launches Issues for Automated Error Monitoring in Workers

Cloudflare launched Issues, a built-in error monitoring tool for Workers that groups production failures and automatically sends diagnostic context to coding agents. The system captures stack traces, logs, and application data, triggering agent workflows to investigate and open pull requests. Issues is now in open beta and requires only a one-line configuration in the wrangler.jsonc file.

Read more
CloudflareCloudflareSep 30

Cloudflare Adds Task, Model, Turn, and User Categories to AI Gateway

Cloudflare updated AI Gateway User Insights to include task, model, turn, and user categories. These additions provide context on AI traffic, helping identify when a model is more capable than a task requires. The update is available free to all AI Gateway users and supports identity-aware tracking for users, teams, and applications.

Read more
CloudflareCloudflareSep 30

Cloudflare Containers Adds Faster Startup, Runtime Configuration, and Snapshots

Cloudflare rearchitected its Containers service for agent workloads, achieving 6x faster startup times. The update introduces a new scheduling policy for runtime selection of sandbox images and instance types. Additionally, filesystem snapshots are now available in public beta, providing persistent workspace states controlled directly from a Durable Object.

Read more
CloudflareCloudflareSep 30

Cloudflare Launches Tools to Identify, Control, and Monetize AI Agents

Cloudflare reports that over half of the traffic on its network is now automated, with AI agents as the fastest-growing segment. To address shifting web economics, the company is launching tools for site owners to identify visitors, set access terms, and charge for content usage through new Pay Per Use and Monetization Gateway features.

Read more
CloudflareCloudflareSep 29

Cloudflare Debuts Adaptive Security Framework for Continuous AI-Era Protection

Cloudflare introduced an adaptive application security framework that links risk discovery, agent governance, runtime protection, and AI-powered response into a continuous learning loop. The system includes new capabilities like LLM-based WAF pentesting and Application Profiles, which automatically learn application structures to block non-conforming requests. Cloudflare also expanded access to its Threat Events Platform to all accounts for free.

Read more
CloudflareCloudflareSep 28

Cloudflare One Adds Security Governance for Model Context Protocol Servers

Cloudflare One now secures connections to Model Context Protocol (MCP) servers, governing AI agent access to internal data. The platform treats MCP servers as managed identities, providing centralized visibility and control over agent interactions with corporate systems. It analyzes prompt intent to block sensitive data leaks, preventing unauthorized connections across the network.

Read more
CloudflareCloudflareSep 28

Cloudflare Updates Kitesurf Browser with WebMCP, DOM Optimizations, and Terminal Rendering

Cloudflare updated Kitesurf, its Workers-based browser for AI agents, adding WebMCP support, improved DOM performance, and terminal-based rendering. The browser now passes over 730,000 Web Platform subtests, allowing agents to navigate complex sites more efficiently. Kitesurf remains available for free during its beta period via the Browser Run API and a new terminal command.

Read more
CloudflareCloudflareSep 28

Cloudflare Applies Zero Trust Policies to NVIDIA OpenShell Agent Traffic

Cloudflare routes NVIDIA OpenShell traffic through its AI Security platform to apply enterprise Zero Trust policies to autonomous agents. This integration governs agent access to the Internet, private applications, MCP servers, and models, enforcing security controls regardless of where the agent runs.

Read more
CloudflareCloudflareSep 16

Cloudflare Client-Side Security ML Detects Payloads Bypassing Traditional Scanners

Cloudflare's Client-Side Security ML detected eight malicious payloads across four operations that bypassed traditional scanners. Seven of the eight payloads had zero detections on VirusTotal, and URLScan flagged none. The system identified threats including affiliate hijacking, invisible iframes, remote backdoors, and analytics silencing by analyzing JavaScript execution patterns rather than relying on static signatures.

Read more
CloudflareCloudflareSep 15

Cloudflare Launches Disallow AI Training Setting for Mixed-Use Crawlers

Cloudflare launched the Disallow AI Training setting to block AI training while maintaining search indexing for mixed-use crawlers. Bot Preference Sync publishes the preference in robots.txt, which Apple, Google, and Microsoft honor or have committed to honor. Existing Block settings now apply to all training crawlers, including mixed-use crawlers, which impacts search discoverability.

Read more
CloudflareCloudflareSep 14

Cloudflare One Adds Governance for Model Context Protocol Server Connections

Cloudflare One now secures connections to Model Context Protocol (MCP) servers, allowing organizations to govern AI agent access within their SASE platform. The update treats MCP servers as managed identities, providing centralized visibility and control over how agents interact with internal data and preventing unauthorized connections across the network.

Read more
CloudflareCloudflareAug 31

Cloudflare Launches Adaptive Intelligence to Undermine Automated Bot Attack Economics

Cloudflare launched Adaptive Intelligence, a bot detection engine that analyzes live traffic meta-signals to generate disposable detection rules that continuously adapt. The engine makes automated attacks economically unsustainable by forcing operators to constantly redesign their bots to match shifting defenses. It activates via the Auto Update Machine Learning toggle in Cloudflare's Bot Management dashboard.

Read more
CloudflareCloudflareAug 22

Cloudflare Launches Bot Preference Sync for Automated robots.txt Management

Cloudflare launched Bot Preference Sync, a feature that automatically aligns a website’s robots.txt file with AI bot preferences configured in the dashboard. By syncing settings for Search, Agent, and Training, the tool eliminates the need to manually maintain static files. The feature rolls out to all customers, from Free to Enterprise, in the coming week.

Read more
CloudflareCloudflareAug 22

Cloudflare OAuth Adds Optional Scopes for Granular User Consent

Cloudflare OAuth now supports optional scopes, enabling client owners to designate specific permissions as optional during configuration. Authorization flows now include a selection step where optional scopes remain unselected to grant a narrower set of access. The resulting access token contains only the approved permissions, providing a more flexible consent experience for agents and internal tools.

Read more
CloudflareCloudflareAug 22

Cloudflare One Now Secures Connections to Model Context Protocol Servers

Cloudflare One now secures connections to Model Context Protocol (MCP) servers, allowing organizations to govern AI agent access as part of their SASE platform. The update treats MCP servers as managed identities, providing centralized visibility and control over how agents interact with internal data and preventing unauthorized connections across the network.

Read more
CloudflareCloudflareAug 14

Cloudflare Gateway Adds MCP Traffic Detection and Security Controls

Cloudflare Gateway now identifies Model Context Protocol traffic using protocol-level heuristics, allowing security teams to detect shadow agent connections. A new dashboard provides visibility into MCP usage, while Gateway policies can now block direct connections to unapproved servers and enforce Portal-only access for managed network paths.

Read more
CloudflareCloudflareAug 8

Cloudflare Launches Radar Researcher for Natural Language Internet Data Exploration

Cloudflare launched Radar Researcher, a beta AI-powered tool that interprets natural language queries to visualize global internet traffic data. The assistant fetches live API data and renders interactive charts rather than text summaries. Radar also adds WebMCP support, providing browser-based AI agents direct access to navigate the platform, scan URLs, and perform domain lookups.

Read more
CloudflareCloudflareAug 8

Cloudflare Unifies Workers AI and AI Gateway Into One Control Plane

Cloudflare merged Workers AI and AI Gateway into a single control plane, providing unified observability, billing, and routing for all AI inference. The update allows Workers AI traffic to route through a default gateway for automatic token tracking and cost attribution. It also enables unified billing across all supported model providers and introduces plans for model-first routing.

Read more
CloudflareCloudflareAug 8

Cloudflare Shifts Bot Mitigation to Continuous Trust Evaluation Framework

Cloudflare is moving bot mitigation from point-in-time risk assessment to continuous trust evaluation. The system uses BotBase to track bot reputation and Precursor to analyze user behavior throughout a session. Cloudflare also announced Adaptive Intelligence, a self-learning detection engine, and upcoming advanced mitigations like AI Labyrinth to manage automated traffic based on intent and data use.

Read more
CloudflareCloudflareAug 6

Cloudflare Adds Stateless MCP Support for Edge-Based Agent Servers

Cloudflare now supports the stateless MCP 2026-07-28 specification, allowing MCP servers to run directly on Cloudflare Workers without stateful infrastructure. The update includes a migration path from the legacy McpAgent to the official TypeScript SDK handler. Early adopters including Sentry, Linear, and Anthropic are already running the new specification in production.

Read more
CloudflareCloudflareAug 6

Cloudflare Launches Kitesurf, a Stateless Browser Optimized for AI Agents

Cloudflare launched Kitesurf, a stateless browser built specifically for AI agents that runs entirely on Cloudflare Workers. By removing human-centric features like extensions and visual rendering, Kitesurf achieves 3–7× better CPU and memory efficiency than Chromium for agentic tasks. It is available for free in beta via the Browser Run API by adding the kitesurf parameter.

Read more
CloudflareCloudflareAug 6

Cloudflare Launches Tools to Measure AI Agent Discoverability and Recommendations

Cloudflare launched Agent Readiness and Answer Engine Optimization (AEO) tools to help site owners measure how AI agents discover and recommend their content. The dashboard provides metrics like citation rates and prominence by probing leading AI models. This release follows Cloudflare’s finding that over half of all web requests now originate from machines rather than human users.

Read more
CloudflareCloudflareAug 6

Cloudflare Launches Developer Preview of WebMCP for Browser AI Agents

Cloudflare launched a developer preview of WebMCP, allowing any website on its network to expose structured tools to browser AI agents. A single dashboard toggle injects a bridge script at the edge, enabling agents to interact with site content without code changes. The preview includes tool packs for C2PA content credentials and site-specific MCP server integration.

Read more
CloudflareCloudflareAug 6

Cloudflare AI Search Simplifies Setup and Adds New Pricing Preview

Cloudflare updated AI Search to automatically compose its underlying primitives, replacing manual configuration. The update adds sitemap-less website discovery, custom domain support for public endpoints, and an EmDash plugin. Cloudflare also shared a pricing preview, where embedding and reranking remain free when using default models from the Workers AI catalog.

Read more
CloudflareCloudflareAug 6

Cloudflare Introduces Open Standards for Agent-Web Cooperation

Cloudflare is building open tools and protocols to make the web readable, discoverable, callable, and payable for AI agents. These standards, including Web Bot Auth and x402, replace current blocking dynamics with cooperative interactions. The initiative provides domain owners with infrastructure to manage agent access and monetization, ensuring the internet remains open rather than siloed.

Read more
CloudflareCloudflareAug 5

Cloudflare Opens Private Beta for WriteGuard Agent Governance Layer

Cloudflare launched WriteGuard, a centralized policy and auditing layer for MCP servers. It classifies agent tools by risk, blocks unauthorized actions, and adds agent attribution to downstream applications. The system is now available in private beta for Cloudflare MCP server portals, providing centralized governance for agent write access without modifying individual server code.

Read more
CloudflareCloudflareAug 5

Cloudflare Launches Identity-Aware AI Gateway and User Insights Analytics

Cloudflare launched Identity-aware AI Gateway in open beta and User Insights in general availability. The gateway integrates with Cloudflare Access to attach verified user identities to AI requests, enabling per-user spend limits. User Insights automatically builds behavioral baselines for every person and agent, flagging anomalous usage patterns that indicate potential insider risk or runaway agents.

Read more
CloudflareCloudflareAug 5

Cloudflare Proposes Agent Access Model to Secure Autonomous AI Agents

Cloudflare introduced the Agent Access Model, a reference architecture for securing task-scoped AI agents. The model uses short-lived, task-scoped credentials and a Trust Ratchet to narrow capabilities during execution. Enforcement occurs at the harness and network layers rather than the prompt, while an Agent Activity Log and Grant Review Loop provide evidence-based policy management for autonomous agent workflows.

Read more
CloudflareCloudflareAug 4

Astro Cuts Open Issues by 85% With Automated AI Subagents

Astro maintainers replaced manual issue verification with isolated AI subagents running in GitHub Actions. This automated pipeline reproduces bugs, diagnoses root causes, and ships preview releases for verification. The implementation reduced the project's open issue count by 85%, bringing the backlog from over 200 to about 30 issues.

Read more
CloudflareCloudflareAug 4

Cloudflare Codex Automates Engineering Standards Enforcement Using AI Agents

Cloudflare created the Cloudflare Codex, a governed set of engineering standards that AI agents retrieve and apply at the point of work. The system uses RFC-style requirements to automatically flag violations in code, technical designs, and incident reports. The AI code reviewer has flagged 230,000 deviations and blocked 16,000 merges, while the spec reviewer has evaluated 600 designs.

Read more
CloudflareCloudflareAug 4

Cloudflare Launches CI SDK for TypeScript Pipelines and Self-Healing Agents

Cloudflare launched a CI SDK that runs CI/CD pipelines natively on its platform using TypeScript instead of YAML. The system integrates with Artifacts and Workflows to provide isolated, sandboxed execution with dependency caching. It includes a self-healing agent harness that uses Workers AI to automatically detect build errors, apply fixes, and push commits for approval.

Read more
CloudflareCloudflareAug 4

Cloudflare Launches Cloudflare Agents Dashboard with Agent Tracing

Cloudflare launched Cloudflare Agents, a unified dashboard for managing and debugging AI agent sessions. The platform introduces agent tracing, which records model calls, tool executions, token usage, and subagent delegations. Sessions can be replayed and execution waterfalls inspected to identify performance bottlenecks. Tracing is free in beta, with paid pricing starting October 1, 2026.

Read more
CloudflareCloudflareAug 4

Cloudflare Introduces Agent Development Lifecycle and Software Factory Primitives

Cloudflare is introducing the Agent Development Lifecycle and a suite of primitives to help agents autonomously manage software delivery. These tools, including @cloudflare/ci and Agent Traces, address the bottleneck where AI-generated code outpaces human review and maintenance. The ADLC provides the infrastructure for software factories to handle planning, testing, deployment, and maintenance without human intervention.

Read more
CloudflareCloudflareAug 4

Cloudflare Boosts Frontier Model Throughput via KV Cache and Weight Compression

Cloudflare optimized its Workers AI platform to serve large frontier models like Kimi and GLM more efficiently. The team implemented FP8 quantization for KV caches, doubling context capacity, and INT4 weight compression to accelerate decode speeds by up to 55%. Additionally, a new integrity check layer protects shared cache pages with less than 1% performance overhead.

Read more
CloudflareCloudflareAug 3

Cloudflare One Adds Intent-Based Prompt Analysis to Prevent Data Leaks

Cloudflare One now analyzes AI prompt content and intent to prevent sensitive corporate data from leaking into public models. The platform blocks risky prompts before they reach external services, securing AI adoption within the SASE architecture. This intent-based filtering provides visibility and control over generative AI usage without requiring manual intervention.

Read more
CloudflareCloudflareAug 3

Cloudflare Releases @cloudflare/computer Runtime to Dynamically Orchestrate Agent Compute

Cloudflare released an early preview of @cloudflare/computer, an open-source agent runtime that dynamically orchestrates between fast isolates and full Linux containers. The package provides a shared filesystem and execution environment, allowing agents to select the most efficient backend for each task. This runtime aims to limit container usage to under 10% of an agent's total workload.

Read more

Frequently asked questions

Cloudflare is Network and security company offering AI inference, edge computing, and developer platform services. HeadsUpAI tracks Cloudflare across the AI ecosystem and curates every significant update — the latest being "Cloudflare Enables WebMCP Shopping Tools for 100,000 Commerce.com Stores" (October 21, 2026) — so you get the whole story in a 30-second read.

The most recent Cloudflare update is "Cloudflare Enables WebMCP Shopping Tools for 100,000 Commerce.com Stores" (October 21, 2026). HeadsUpAI curates every significant Cloudflare release as a 30-second read — what shipped and why it matters.

The latest Cloudflare updates: "Cloudflare Enables WebMCP Shopping Tools for 100,000 Commerce.com Stores", "Cloudflare Partners With Mastercard on Agent Pay for Machines", "Cloudflare API MCP Server Adds Step-Up Authorization Support", "Cloudflare Cuts Internal AI Code Reviewer Costs by 68 Percent", and "Cloudflare Launches Clairic, a System One Extraction Model for Typed Values". HeadsUpAI has curated 51 Cloudflare updates over the last 90 days, covering product updates, launches, and company news — listed newest first, presented straight, no hype, no bias.

Cloudflare is Network and security company offering AI inference, edge computing, and developer platform services. On this page you'll find every significant Cloudflare development HeadsUpAI has tracked recently — product updates, launches, and company news — so you can keep up with where Cloudflare is heading without reading a dozen sources.

Continuously. HeadsUpAI adds new Cloudflare updates as they're announced — usually within hours — and the 51 updates currently shown cover the past 90 days, newest first.