Agentic Web: standards, search, and edge agents

OpenAI introduces Agent Plugins open standard for bundling skills and MCP servers; steering committee includes Amazon, Microsoft, Vercel. OpenAI publishes an open Agent Plugins standard and steers governance with industry partners so teams can bundle skills and run them on MCP servers; this shifts how you package, deploy, and govern agent capabilities — Principle 09.

Building an open Agentic Internet: readable, discoverable, callable, and payable. Cloudflare lays out Web Bot Auth, x402, MCP, and PACT to make sites machine-readable and callable by agents, giving outcome engineers a blueprint for interoperable discovery, identity, and payment on the agentic web — Principles 01 and 06.

Cloudflare AI Search: give your agents a search engine for your data. Cloudflare ships AI Search to provide semantic, ingestible search endpoints for agents across sites and docs, making contextual grounding and retrieval a first-class plumbing for agent workflows — Principles 06 and 11.

No cloud, no GPUs, no problem: Liquid AI’s LFM2.5-2.6B brings powerful AI agents to devices as small as a Raspberry Pi. Liquid AI releases LFM2.5–2.6B to run agentic workloads on phones and tiny devices, enabling private, low-latency edge agents and forcing design trade-offs around model size, state, and orchestration at the edge — Principles 04 and 07.

Give any website a WebMCP interface. Cloudflare makes it trivial to expose WebMCP on existing sites so agents can interact without scraping or custom code, which directly reduces integration friction and clarifies how you should expose machine-readable endpoints for outcome-driven agent tasks — Principles 06 and 11.