Everything we launched during Agents Week

Summary

Cloudflare wrapped up its Agents Week by outlining infrastructure and security capabilities intended to support an emerging Agentic Internet, in which autonomous software agents act on behalf of users across web services and internal systems. The company framed agents as a distinct class of software whose requirements extend past model performance to include identity, orchestration, observability, and controlled execution.

Central to the security announcements is the Agent Access Model, which applies Zero Trust principles to agent-initiated requests so that resources can be reached only under explicit, attributable permissions. Identity-aware analytics complement this by linking observed activity to specific users or systems, allowing operators to surface anomalies and unexpected spending patterns that may indicate rogue behavior. WriteGuard adds another layer by supplying fine-grained policy controls over tool calls, particularly for MCP servers, reducing the risk of unintended changes while preserving the same mechanisms Cloudflare uses internally.

These controls sit alongside new runtime primitives, an Agent Development Lifecycle, and support for mixed-language Workers, all presented as foundations for building observable and scalable agent applications. The week’s releases collectively aim to let organizations run agents with traceable actions, programmable transaction capabilities, and enforceable boundaries rather than relying on ad-hoc safeguards.

Why it matters

This article is highly relevant for security and privacy professionals as it introduces practical frameworks and tools, such as Zero Trust for AI agents and identity-aware analytics, to secure autonomous AI systems. These capabilities are crucial for Dutch enterprises aiming to deploy AI agents securely while complying with strict EU data protection and AI regulations.