Deep-Dives / Protocols & Interop

Protocols & Interop

The protocol layer of the agent stack — MCP, A2A, tool-calling standards, capability discovery — and how systems plug together.

  1. Why Interop Matters: The M×N Problem
    How connecting M agents to N systems by hand explodes, and why a protocol layer is the structural fix.
  2. Tool Calling Standards: JSON Schema
    The universal declare/select/execute/return contract, the portable JSON Schema core, and where providers differ.
  3. MCP: Hosts, Clients, Servers
    The Model Context Protocol participant model, resources/tools/prompts, JSON-RPC lifecycle, and transports.
  4. Agent-to-Agent Communication
    Delegating to opaque peer agents: Agent Cards, tasks, messages, artifacts, and long-running work.
  5. Structured Tool I/O & Validation
    Input and output as two trust boundaries: structural-then-semantic validation, and why typed output is still untrusted.
  6. Capability Discovery & Negotiation
    Runtime discovery, feature-test version negotiation, and why discovery describes ability not permission.
  7. Building an Interoperable Agent
    Comparing tool calling, MCP, and A2A; a decision rule and one normalised registry architecture.
  8. A2A v1.0: Task Lifecycle, Messages, Artifacts
    A2A hit v1.0 in April 2026 — nine task states (not four), Message vs Artifact split, A2A-Version header, breaking changes from pre-1.0, and 150+ org adoption.
  9. Agent Cards & Discovery
    A2A's /.well-known/agent.json — capability declaration, extended cards, signing, caching, and how it compares with MCP registry-based discovery.
  10. ACP: What Happened
    A short post-mortem — ACP existed, was REST-native, was contributed to the Linux Foundation in July 2025, and folded into A2A. Useful because search still surfaces stale 'ACP vs A2A' content.
  11. AP2 & Agent Commerce
    The Agent Payments Protocol — Intent / Cart / Payment as W3C VCs, why it sits above A2A/MCP rather than inside them, and the stablecoin-rail pitch to keep skeptical of.
  12. agents.json & OpenAPI for Agents
    The agents.json v0.1 spec on top of OpenAPI, the AGENTS.md convention adopted by 20k+ repos, and why "just point the agent at your OpenAPI" does not fully work.
  13. ACP: The Agent Client Protocol
    Zed's editor-to-agent protocol — the initialize capability gate, the prompt turn, and the filesystem inversion that makes the client, not the agent, hold the disk.
  14. Poppy: The Personal Agent Protocol
    Sierra and Meta's protocol for someone else's agent signing in as your customer — the well-known file that points the other way, the session that starts with no user in it, two reserved scopes that cannot express risk, and the per-company pseudonym that deletes a key you were using.
  15. Web Bot Auth: Signing an Agent’s Requests
    A signature proves which operator's software sent the request and nothing else — not which user, not the errand, not a scope — so it retires impersonation and leaves delegation untouched. Adopted as the webbotauth working group's first draft in September 2026 over RFC 9421, which is the stable half; verification spread much faster than signing (Cloudflare's Verified Bots Program, Akamai, AWS WAF, HUMAN, Vercel, Stytch against Google, OpenAI and Bedrock AgentCore), so a signature-required rule still blocks most legitimate agents. Three buckets, not two — and never let a verified operator name reach where a user identity is checked.