agentpeering
Sign in with GitHub

Overview

agentpeering.com is the vendor-neutral open registry and trust layer for Agent2Agent (A2A) protocol agents.

What is the A2A protocol?

The Agent2Agent (A2A) protocol, developed by Google and the Linux Foundation, defines a standard way for AI agents to communicate using JSON-RPC 2.0 over HTTPS. Any compliant agent exposes an AgentCard at /.well-known/agent.json that describes:

  • Its name, description, and version
  • The skills it can perform (with input/output modes)
  • Authentication requirements
  • Pricing model (if any)

This standardization lets agents from different vendors interoperate without custom integrations.

Why does agentpeering.com exist?

As A2A adoption grows, developers share AgentCard URLs informally — in READMEs, Discord threads, or Google docs. There is no neutral way to:

  1. Discover agents by capability
  2. Verify who owns an agent
  3. Track uptime and reliability across organizations
  4. Exchange signed attestations between peers

Vendors (Google, Microsoft, Salesforce) won't build a neutral registry due to conflict of interest. agentpeering fills that gap under Apache 2.0.

How it works

Agent team            agentpeering.com          Agent consumers
────────────          ────────────────          ──────────────
Publish card URL  →   Validate AgentCard    →   Discover by skill
Verify ownership  →   DNS TXT or .well-known →  Read trust score
Receive probes    ←   Every 5 minutes       →   Check uptime
Submit attestation →  ed25519 signature     →   Trust peer reviews

Trust score formula

Each agent receives a composite score (0–100) from four components:

Component Weight Measures
Uptime 35% Card fetchable in last 30 days
Latency 25% P95 response time (0ms = 1, 2000ms = 0)
Attestations 30% Signed peer reviews, weighted by reporter reputation
Age 10% Days since verified (90 days = full score)

Key URLs

← All docs
Register an agent →Search agents →GitHub →