OpenAI pushes agent execution with WebMCP standard and rising enterprise token usage

Site tools are ChatGPT’s implementation of the proposed WebMCP standard. [1] WebMCP is an experimental open standard that lets web apps expose tools for agents to use directly on the page. [1] In the built-in browser in the ChatGPT desktop app, ChatGPT Work and Codex can discover and use these tools when they are available. [1]

To promote the standard, OpenAI teamed up with Google Chrome, Cloudflare, Shopify, Vercel, Render, and Netlify for a 10-day hackathon offering $35,000 in cash prizes. [1]

Frontier firms in the top ten percent of AI usage now generate 8.3 times as many output tokens per active user as typical firms. [2] This token generation gap is up from a 2.6 times multiple in January. [2]

Basis, a startup that builds AI agents for accounting firms, reduced its first-day onboarding time from two hours to 30 minutes. [2] On day one, employees receive immediate access to Codex, which introduces key company concepts and uses their computer to complete integration setup in the background. [2]

These developments demonstrate OpenAI's dual approach to agentic execution: building open standards for external web interaction while accelerating internal enterprise workflows. The shift from simple assistance to execution is evident in both the surging token usage among top enterprise adopters and the push to standardize how web applications communicate with models like Codex. By partnering with major platforms like Google Chrome and Cloudflare for WebMCP, OpenAI is attempting to establish the fundamental architecture for agent-to-web interactions before proprietary silos take hold. What the sources don't address: How developers will secure authenticated WebMCP sessions against unintended agent actions or excessive resource consumption.