WebMCP & Me
How doesMCP & wMCP give AI the keys to the web ?
If you are building or experimenting with AI agents right now, you know the frustration.
We have these incredibly smart models that can write code and reason through complex logic, but the second you ask them to interact with the actual internet ex. checking a product price, comparing variants, or adding an item to a cart, they hit a wall.
They either hallucinate, get blocked by bot protection like Cloudflare or Akamai, or require you to spend hours writing custom, brittle scrapers for every single website. Every framework reinvents the same scraping logic, and the same Shopify cart endpoints get rediscovered weekly.
It shouldn't be this hard to let an agent browse a store. That is exactly the problem wmcp.sh solves.
What is wmcp.sh?
Think of it as a universal translation layer for the web, built on the Model Context Protocol (MCP).
Instead of writing a custom integration for every site, you just hand a URL to wmcp.sh. It instantly returns a typed, ready-to-use list of tools your AI can understand and call.
• Shopper-Side, Not Admin-Side: Most e-commerce AI tools are built for merchants and require admin API keys. WebMCP is built for the shopper. It works on public storefronts (over 4 million Shopify stores, JSON-LD sites, and OpenAPI specs) without needing any credentials.
• Drop-In Integration: Whether you are using the Anthropic SDK, OpenAI function calling, LangChain, or Cursor, it takes about three lines of code to drop it into your stack. The AI gets the tools, decides what to call, and the call returns real data.
• The Bot-Block Bypass: This is the killer feature. For heavily protected sites that block standard server requests (like Nike or Amazon), the WebMCP Chrome extension extracts the schema client-side in a real browser.
The Network Effect
Here is the best part: it gets faster the more people use it.
When that Chrome extension extracts a schema from a blocked site, it pushes it back to a shared global cache. The next time anyone's agent asks for that same URL, the cached schema is served instantly in under 100ms. Every install makes the API better for everyone else, even for users who never install the extension themselves.
We are moving past the point of chatbots just talking to us. We need them to act.
If you are tired of your agents getting bounced by basic web firewalls or want to stop writing custom scrapers, head over to wmcp.sh. Paste a product URL into the live demo, see the tools it generates, and give your AI the access it’s been waiting for.
Please subscribe to follow along as I dive deeper into the WebMCP ecosystem, share new adapters, and document what it takes to build AI that actually works in the real world.
