Introduction to the Agent eXchange Layer (AXL)

on YouTube ↗15m

" summary

A peer‑to‑peer mesh network implemented as a single Go binary handles routing, NAT traversal, hop‑by‑hop TLS plus end‑to‑end payload encryption, and node identity via public keys; no central servers, DNS, port forwarding, or cloud accounts are required. Applications call three HTTP endpoints on localhost (default port 9002): send, receive, topology. Supports MCP and Google A2A for named services and skills. Demo shows two nodes running a group chat and a shared OpenClaw agent that preserves conversation context across users, and bundled examples include gossip, tensor exchange, and pipeline parallelism.

AI-written summary · about

" chapters (23)

1. 0:00Introduction
2. 0:12What Is the Agent eXchange Layer (AXL)
3. 0:53How AXL Works: Peer-to-Peer Mesh Network
4. 1:36Application Agnostic: Any Language, Any Payload
5. 1:57Public Keys as Node Identity
6. 2:17No Central Servers, No Port Forwarding
7. 2:41Three Core HTTP Endpoints: Send, Receive, Topology
8. 3:22Dual-Layer Encryption: TLS and End-to-End
9. 4:26Building with AXL: Send and Receive Pattern
10. 5:15Use Cases: Agent Collaboration and Chat Rooms
11. 5:40MCP and A2A Protocol Support
12. 6:32Getting Started: Clone, Build, Generate Identity
13. 7:12What You Can Build with AXL
14. 7:29Agent Collaboration: Decentralized Research
15. 7:48Service Registry and Tool Marketplace
16. 8:23Private Data Exchange with Encryption
17. 8:51Distributed Inference and Pipeline Parallelism
18. 9:17Packaged Examples: Gossip, Tensor Exchange, A2A
19. 9:58Live Demo: Peer-to-Peer Group Chat with OpenClaw Agent
20. 11:27Shared Agent Instance Across Multiple Users
21. 13:45How the Demo Works Under the Hood
22. 14:23Resources and Getting Started
23. 14:41Closing

" related talks

• 19mSecuring Agents with Open StandardsBobby Tiernay, Kam Sween · AI Engineer World's Fair 2025
• 23mWho is Calling? Identity for Agents with AAuthDick Hardt · Confidential Computing Summit 2026
• 33mAI  Your DataWilliam Blankenship · HalfStack Phoenix 2025
• 01h 22mIdentity for AI AgentsPatrick Riley, Carlos Galan · AI Engineer CODE 2025
• 24mWhat if the network was the sandbox?Remy Guercio · AI Engineer Europe 2026
• 40mIn AI We Trust? Securing the Future, One Agent at a TimeRicardo Aravena, Lin Sun, Eitan Yarmush, Hannah Foxwell · Open Source SecurityCon North America 2025
• 54mMaking oRAT Go...FurtherPatrick Wardle · RSA Conference 2023
• 45mHow can applications become more secure, private, and dependableMrinal Wadhwa · Alchemy Conf 2021