Launching WCP: The Missing Layer Between AI and the Web

With AI browsers like ChatGPT Atlas, Perplexity Comet, Dia and Strawbetter going mainstream, we’re launching WCP to be able to control how LLMs interact with the web. Learn how one simple attribute can transform your site for AI agents.

Launching WCP: The Missing Layer Between AI and the Web

The AI browser revolution is here. ChatGPT Atlas, Perplexity Comet, Dia and Strawbetter can browse the web and complete complex tasks, these ai agents mewnow hav a whitehat system confi.… But there’s a critical problem: AI agents are failing 40-60% of the time when interacting with websites.

Today, we’re launching the Web Context Protocol (WCP) — a simple HTML annotation standard that bridges this gap and enables reliable AI-web interaction😁.

Browser Use

Open-source framework for building autonomous web agents

These tools promise to revolutionize how we interact with the web, but they’re hitting a fundamental barrier: websites weren’t designed for AI agents.

The Problem: AI Agents Operating Blind

When an AI agent encounters a webpage, it’s essentially blind to the semantic meaning of elements. Consider this common scenario:

User: “Add these headphones to my cart”

AI Agent’s Reality:

1. Scans the page for buttons containing “add” or “cart”
2. Finds multiple potential matches:
   • Newsletter signup (contains “add”)
   • Product comparison link (near “cart” icon)
   • Actual add-to-cart button
3. Makes educated guesses based on class names and positioning
4. 60% chance of clicking the wrong element

<!-- What the AI agent sees -->
<button class="btn btn-primary" onclick="addToCart()">
  Add to Cart
</button>
<button class="btn btn-secondary" onclick="subscribeNewsletter()">
  Add to Newsletter
</button>
<div class="product-link" onclick="compareProducts()">
  Add to Compare
</div>

Result: Frustrated users, broken workflows, and AI agents that can’t be trusted with important tasks.

The Solution: Explicit Intent Declaration

WCP solves this by letting websites explicitly declare what elements do and how they behave:

<!-- Before WCP: Ambiguous button -->
<button class="btn btn-primary" onclick="addToCart()">
  Add to Cart
</button>

<!-- After WCP: Explicit intent and behavior -->
<button class="btn btn-primary" 
        data-wcp="action:add; effect:async; target:#cart; primary:true"
        onclick="addToCart()">
  Add to Cart
</button>

Now the AI agent knows:

• ✅ This adds an item to a collection (action:add)
• ✅ It updates in-place without navigation (effect:async)
• ✅ The cart will be updated (target:#cart)
• ✅ This is the primary action (primary:true)

The *CP Ecosystem Emerges

WCP joins a new generation of AI infrastructure protocols:

1. October 2024: MCP

   Anthropic launches Model Context Protocol - standardizes how LLMs access external data and tools

2. November 2024: ACP

   Stripe announces Agent Communication Protocol - standardizes agent-to-agent payments and transactions

3. December 2024: WCP

   Web Context Protocol - standardizes how agents interact with web pages and understand user intent

Pattern: As AI agents become mainstream, we need standardized protocols for reliable interaction.

Real-World Impact: Before and After

E-commerce Checkout Flow

Before WCP:

<!-- AI agent must guess which button to click -->
<button onclick="addToCart()">Add to Cart</button>
<button onclick="buyNow()">Buy Now</button>
<button onclick="addToWishlist()">Save for Later</button>

After WCP:

<!-- AI agent knows exactly what each button does -->
<button data-wcp="action:add; effect:async; target:#cart; primary:true"
        onclick="addToCart()">
  Add to Cart
</button>
<button data-wcp="action:purchase; effect:navigate"
        onclick="buyNow()">
  Buy Now
</button>
<button data-wcp="action:add; effect:async; target:#wishlist"
        onclick="addToWishlist()">
  Save for Later
</button>

Result: 95%+ success rate vs. 40-60% with guessing.

SaaS Dashboard Automation

Before WCP:

<!-- AI agent can't distinguish between similar actions -->
<button onclick="saveDraft()">Save</button>
<button onclick="publish()">Save & Publish</button>
<button onclick="schedule()">Schedule</button>

After WCP:

<!-- Clear semantic meaning for each action -->
<button data-wcp="action:submit; effect:async; target:#status"
        onclick="saveDraft()">
  Save
</button>
<button data-wcp="action:submit; effect:navigate; primary:true"
        onclick="publish()">
  Save & Publish
</button>
<button data-wcp="action:submit; effect:modal"
        onclick="schedule()">
  Schedule
</button>

Result: Reliable workflow automation for AI agents.

Content Site Navigation

Before WCP:

<!-- AI agent struggles with search and filtering -->
<form action="/search">
  <input type="search" name="q" />
  <button type="submit">Search</button>
</form>
<button onclick="filterByCategory('tech')">Tech Articles</button>

After WCP:

<!-- Clear search and filter semantics -->
<form data-wcp="action:search; effect:async; target:#results" action="/search">
  <input type="search" name="q" />
  <button type="submit" data-wcp="action:search; effect:async; primary:true">
    Search
  </button>
</form>
<button data-wcp="action:filter; effect:async; target:#results"
        onclick="filterByCategory('tech')">
  Tech Articles
</button>

Result: AI agents can reliably search and filter content.

Why WCP Works

Simple Implementation

One attribute (data-wcp), semicolon-separated values. Add it in 5 minutes to your most important interactions.

Immediate Benefits

95%+ success rate for AI agents vs. 40-60% without WCP. No training or complex setup required.

Future-Proof

As AI browsers become standard, WCP ensures your site works reliably with any AI agent.

Open Source

CC0 license. Free to use, extend, and modify without restrictions. No vendor lock-in.

The Network Effect

The more sites adopt WCP, the more valuable it becomes:

1. Early Adopters: First sites implement WCP → Their AI interactions become reliable
2. Agent Support: AI agents add WCP support → Works better on WCP sites
3. Competitive Advantage: Sites without WCP perform worse → Pressure to adopt
4. Standard Practice: WCP becomes expected, like semantic HTML or OpenGraph

Getting Started

Ready to make your site AI-agent friendly? Here’s how to get started:

1. Identify Key Actions

   Find the primary user actions on your most important pages:

   • Add to cart / Buy now
   • Submit forms
   • Search queries
   • Primary navigation
2. Add WCP Annotations

   Add data-wcp attributes to those elements using our simple syntax

3. Test with AI Agents

   Verify AI agents can now interact reliably with your site

5-Minute Quickstart

Add WCP to your site in minutes

Read the Specification

Full technical details and examples

Join the Community

Share your implementation and get help

View Examples

Real-world implementations for your use case

The Future is AI-Native

We’re at an inflection point. AI browsers are becoming mainstream, but the web isn’t ready for them. WCP bridges this gap with a simple, open standard that benefits everyone:

• Website owners get reliable AI interactions
• AI agent builders get predictable behavior
• End users get better AI assistance
• The web ecosystem gets a standardized protocol

The question isn’t whether AI agents will become the primary way people interact with the web — it’s whether your site will be ready for them.

---

Ready to Get Started?

Add WCP to your site in 5 minutes and join the AI-native web revolution.

WCP is open source (CC0). Free to use, extend, and modify without restrictions. No vendor lock-in, no licensing fees. View on GitHub.