Today, most web browsing AI agents rely on DOM and screenshot analysis, and then guessing which clicks to make. WebMCP is the bridge from "AI can read pages" to "AI can safely and predictably do work on pages" by replacing guesswork with actual web developer involvement and explicit tool contracts. In this talk I show how WebMCP turns multi-step, high-friction workflows into structured tool calls you can trust. WebMCP eliminates DOM brittleness, is resilient to UI churn, doesn't require shipping a separate server-side MCP endpoint, and let's you incrementally make your existing site agent-ready.
Video, slides.
Current
·
C2
·2026-09
WebMCP: Stop scraping the web, start calling it (dotAI, Paris)
Find Your Interest
- I'm worried about the web
- AI agents and the web - A proposal to keep developers in the loop
- How AI-assisted development helped us build the new Dropbox experience
- W3C and the State of the Web in the AI Era
- The Agentic Web
- Bringing Goodnotes to the web with Swift and WebAssembly
- CSS Zen Posters
- WebMCP updates, clarifications, and next steps
- Introducing the Heroku AI Platform as a Service (AI PaaS)
- Heroku: Powering the Next Wave of Apps with AI
Current
·
C2
·2026-09






