OpenAI Adds WebMCP Site Tools to ChatGPT Browser
OpenAI has added WebMCP support to the built-in browser in its ChatGPT desktop app, letting websites expose actions that ChatGPT and Codex can call directly from the page. The feature is separate from the server-based Model Context Protocol that ChatGPT has supported since 2025, and does not require a separate connector to be set up.
A webpage registers JavaScript functions as tools, including names, descriptions and structured input schemas, so an agent can use an app's own logic and permissions rather than guessing its way through the interface. An arrow in the address bar shows when site tools are available and whether a tool can read data or make changes. Tools stay tied to the page they were accessed from and disappear when it closes.
OpenAI describes WebMCP as an experimental open standard. The specification is a draft from the W3C Web Machine Learning Community Group and is not on the W3C Standards Track. Site tools require GPT-5.6 Sol or Terra, are disabled on GPT-5.6 Luna, and are not available in Enterprise or Edu workspaces. Developers can test the standard in Chrome through an experimental flag or its origin trial.
For brands trying to be cited by AI assistants, the feature gives site owners a way to decide which actions an agent can take, rather than leaving that to ChatGPT's default clicking and typing. OpenAI's documentation does not say how WebMCP affects rankings, citations or discoverability. The company warns of data exfiltration and prompt injection risks, and asks for permission before an agent interacts with a site or confirms actions such as purchases, deletions or messages.
Reported by Search Engine Journal ↗
Topics: ChatGPT, Search policy