OpenAI shows how to build custom site annotations with WebMCP for ChatGPT
OpenAI · youtube · 2026-10-08
OpenAI published a tutorial showing developers how to build custom site annotations for ChatGPT using WebMCP. Websites can add custom annotation controls and suggested prompts that appear when users browse with ChatGPT's agent. Two demos: letting users edit an on-page scene via ChatGPT, and enriching expense records with context available to the agent. It's a concrete look at exposing site capabilities directly to ChatGPT users through the WebMCP protocol.
More from coding & agent
- ChatGPT App Controlling Claude Controlling ChatGPT: An Agent-Calling-Agent Moment — danshipper · 2026-10-08
- ML Author Andriy Burkov Mulls Writing a Hands-On 'Hundred-Page Agentic Harness Book' in Python — burkov · 2026-10-08
- Every killed its personal Slack agents and bet on one shared company agent — danshipper · 2026-10-08
- Most AI coding slop comes from being too complacent with agents, says JPMorgan quant researcher — JFPuget · 2026-10-08
- Agent evals 101: record full runs, label pass/fail, skip 1-5 scores — strickvl · 2026-10-08
- Fail an AI agent twice, add regression tests: a practical agent debugging workflow — strickvl · 2026-10-08