Connect via MCP

This site supports WebMCP, which means your AI agent can interact with it directly. Search posts, read content, and explore the blog without screenshots or scraping.

How to Connect

  1. 1. Install the WebMCP MCP server

    Run this in your terminal (picks up your MCP client automatically):

    npx -y @jason.today/webmcp@latest --config claude

    Replace claude with cursor, cline, or windsurf if you use a different client.

  2. 2. Generate a token

    Ask your AI agent to generate a WebMCP token, or run:

    npx @jason.today/webmcp --new
  3. 3. Paste the token

    Click the widget in the corner of this page and paste your token. Once connected, your agent has access to the tools below.

Available Tools

search_posts

Search blog posts by keyword. Returns matching titles, slugs, and descriptions.

get_post

Get the full content of a blog post by its slug.

list_posts

List all published blog posts with titles, dates, and tags.

get_site_info

Get information about this site, the author, and what topics are covered.

What is WebMCP?

There are two things called "WebMCP" right now, and they solve the same problem differently.

jasonjmcghee/WebMCP (what this site uses)

An open source JavaScript library that lets any website expose MCP tools to AI agents via a local WebSocket bridge. Works today with Claude Desktop, Cursor, and any MCP client. No special browser required.

GitHub · npm

W3C/Chrome WebMCP (the browser spec)

A proposed W3C web standard backed by Google Chrome and Microsoft Edge that adds a native navigator.modelContext API to browsers. Currently in early preview behind a flag in Chrome Beta (146+). Not yet in stable browsers.

MCP Spec · W3C Proposal