Cloudflare

Cloudflare has launched its Clef and Clef-flash models, specifically designed for rapid, autonomous AI agent decision-making. These open-weight models, built on the Qwen architecture, excel at classification and routing tasks by returning typed probabilities rather than generating free-form text. Clef-flash, in particular, boasts impressive speed, completing classifications in under 40 milliseconds, positioning Cloudflare as a key enabler for efficient AI agent workflows. Cloudflare is significantly enhancing its infrastructure to support the burgeoning AI agent ecosystem, from data storage to API access. New offerings like Polylinedb, running on Cloudflare Workers and D1, provide unified task and memory stores for cloud agents. The company also launched an AI Web Search API, integrating services like Exa and Ceramic, and is actively promoting the Model Context Protocol (MCP) for streamlined tool integration, solidifying its role as a foundational layer for agentic AI. Cloudflare is introducing new mechanisms for monetizing AI agent interactions and data access, including token-based pricing and programmable wallets. The upcoming token-based pricing for Cloudflare AI Search, starting November 1st, will charge per million tokens, providing a clear cost structure for AI-driven queries. Additionally, the launch of programmable wallets for AI agents and a Monetization Gateway with USDC settlement enables agents to autonomously pay for services, fostering a new economy for digital interactions. Cloudflare is actively bolstering its security posture with AI-powered tools and grappling with the ethical implications of its infrastructure. The company open-sourced AI bug-hunting agents to proactively identify vulnerabilities in its code, demonstrating a commitment to secure development. Concurrently, Cloudflare is refining its Web Application Firewall using frontier AI models for advanced threat detection. However, a recent study highlighted its inadvertent role in hosting abusive deepfake content, prompting ongoing scrutiny regarding platform accountability. Beyond AI-specific tools, Cloudflare introduced K2 serverless streams for enhanced data processing and EmDash, an open-source CMS for humans and AI agents. The company is also incentivizing developers with $25,000 in credits to build a Git-like version control system for AI agents on Cloudflare Workers and Artifacts, signaling a push towards more robust and collaborative AI development environments.