A2A

This section is advisory.

A2A is an open protocol (Apache 2.0) enabling communication between AI agents. Created by Google, now governed by the Agentic AI Foundation (AAIF) under the Linux Foundation, with nearly 150 member organizations. Version 1.0 supports JSON-RPC, gRPC, and HTTP/REST bindings.

Agent Card Extension — Publishing Legal Requirements

A2A agents discover each other via Agent Cards published at /.well-known/agent-card.json. An LCP-aware agent can declare its legal context requirements in its Agent Card, enabling programmatic compatibility assessment before interaction begins:

{
  "name": "Procurement Agent",
  "skills": [...],
  "legalContext": {
    "required": true,
    "minimumLevel": 2,
    "acceptedJurisdictions": ["New York, USA", "London, UK"],
    "acceptedDisputeMethods": ["AAA Commercial Arbitration Rules", "ICC Rules"]
  }
}

This allows agents to evaluate legal compatibility during discovery — before any transaction begins.

Task-Level Legal Context

Before agents collaborate on a task via A2A, the initiating agent can include LCP references in the task metadata, establishing the legal framework governing the collaboration:

{
  "task": {
    "id": "task-123",
    "metadata": {
      "legalContext": {
        "type": "sha256",
        "value": "0x7f83b165..."
      }
    }
  }
}

This makes the legal context available throughout the task lifecycle.

Skill-Level Terms

Each agent skill can declare required LCP terms, enabling agents to negotiate legal context as part of capability negotiation rather than as an afterthought. When an agent advertises skills that involve legally significant actions — procurement, contracting, payments — the skill definition can specify the LCP terms that must be accepted before invocation.

Integration Flow

1. Agent discovers a counterparty via its Agent Card at /.well-known/agent-card.json
2. Agent evaluates the legalContext requirements — jurisdiction, minimum level, accepted dispute methods
3. If requirements are compatible, agent initiates a Task with legalContext in task metadata
4. During the task, skill-level terms are negotiated as needed
5. Legal context reference travels with the task throughout the collaboration lifecycle