World Agent Web – Protocols for AI Agents to Discover and Use Web Services
Key topics
The Problem: Today, AI agents access web services through either: - Web search + HTML scraping (fragile, breaks when sites change) - Hardcoded custom integrations (doesn't scale)
Neither approach works well. Agents need a standardized way to discover what services can do and how to invoke them.
The Solution: WAW introduces manifest-based discovery (similar to robots.txt, but for agents): - Services publish a manifest describing their capabilities - Agents discover these dynamically at runtime - Type-safe invocation with JSON Schema validation - No hardcoded integrations needed
Think of it as: HTTP is to the World Wide Web what WAW is to the Agent Web.
Live Demo: - Working services: https://web-production-73c2.up.railway.app - Autonomous Claude agent: https://github.com/Fredbcx/waw/tree/main/demo - The agent discovers services, understands capabilities, and books restaurants autonomously
Positioning: WAW complements existing protocols: - MCP (Model Context Protocol): Desktop/local tool integration - A2A (Agent-to-Agent): Multi-agent collaboration - WAW: Web service discovery and invocation
They solve different problems and work together.
What I'm looking for feedback on: 1. Protocol design - does the spec make sense? 2. Use cases I'm missing 3. Integration patterns with existing systems 4. Whether this actually solves a real problem you have
The protocol is MIT licensed and designed to be community-driven. I see service registries (like Google for agent services) and micropayments (x402) as inevitable future infrastructure, but wanted to validate the core protocol first.
Links: - GitHub: https://github.com/Fredbcx/waw - Live site: https://web-production-73c2.up.railway.app - Spec: https://github.com/Fredbcx/waw/blob/main/docs/SPECIFICATION.md
Would love your thoughts!
The World Agent Web project proposes protocols for AI agents to discover and use web services, sparking interest and discussion about its potential and comparison to existing approaches like function calling.
Snapshot generated from the HN discussion
Discussion Activity
Light discussionFirst comment
34m
Peak period
2
0-3h
Avg / period
1.3
Key moments
- 01Story posted
Nov 11, 2025 at 12:00 PM EST
about 2 months ago
Step 01 - 02First comment
Nov 11, 2025 at 12:34 PM EST
34m after posting
Step 02 - 03Peak activity
2 comments in 0-3h
Hottest window of the conversation
Step 03 - 04Latest activity
Nov 13, 2025 at 6:24 AM EST
about 2 months ago
Step 04
Generating AI Summary...
Analyzing up to 500 comments to identify key contributors and discussion patterns
Want the full context?
Jump to the original sources
Read the primary article or dive into the live Hacker News thread when you're ready.
Think of it as: - Function calling: Client tells LLM "here are tools you can use" - WAW: Service tells agent "here's what I can do"
You can actually combine them—use WAW manifests to auto-generate function definitions dynamically. We show this in the demo.