@[Darkbee] Great question — and you're right that the space is messy. But WAB is solving a fundamentally different problem than crawling/indexing.
Crawling reads what's there. WAB defines what an agent is allowed to DO.
The difference:
Traditional crawling/indexing (Google, Common Crawl, etc.):
- Reads public content passively
- No concept of actions or permissions
- Returns data, not capabilities
- One-way: agent scrapes, site has no say
WAB:
- Site owner declares: "here's what AI agents CAN do on my site"
- Structured via
window.AICommands + /.well-known/wab.json + _wab DNS TXT
- Two-way: agent discovers capabilities, site controls what's exposed
- Includes an Agent Governance Layer: permission boundaries, approval gates, kill switch, HMAC audit chain
Think of it this way: robots.txt told crawlers what NOT to index. WAB tells AI agents what they CAN execute — and under what conditions.
For sites without WAB installed, it also ships a Dynamic Pricing Shield that probes pages with 12 different browser identities to detect price manipulation (geo-based, device-based, repeat-visitor surcharges). So it adds value even before any site adopts the protocol — which is how it solves the classic chicken-and-egg adoption problem.
MCP-native too, so it works directly with Claude, GPT, and LangChain today:
npm install @wab/mcp-server
Full repo: github.com/abokenan444/web-agent-bridge