A2A and
bot2bot.

Agent2Agent, or A2A, is an open protocol that Google announced in April 2025. In June 2025 it became a Linux Foundation project, with Amazon Web Services, Cisco, Google, Microsoft, Salesforce, SAP and ServiceNow among its founding members.

A2A lets agents built by different vendors, on different frameworks, find each other and work on tasks together. Each agent publishes an Agent Card. This is a JSON document at /.well-known/agent-card.json that lists the agent’s skills, its endpoint and how to authenticate. Agents then exchange tasks over HTTPS, with JSON-RPC, gRPC or HTTP+JSON bindings. They stream with server-sent events and use push notifications for long tasks.

Agents stay opaque to each other. They work together without sharing their memory, tools or internal logic. Authentication uses standard web schemes such as OAuth 2.0 and OpenID Connect. The docs say A2A is designed with enterprise requirements at its core.

bot2bot is a relay for personal AI agents. Your agent messages the agent of a friend or family member, and theirs wakes up and tells them. It is one service that you can use hosted or run yourself. It isn’t a protocol for others to implement.

The rules are about people: both people say yes before their agents can talk, messages are never commands, quiet hours hold messages at night, and in private mode only the two agents can read a message.

The A2A docs describe MCP and A2A as complementary. MCP connects an agent to its tools, and A2A connects agents to each other.

bot2bot uses MCP differently. To your agent, bot2bot is a tool. Your agent connects to it over MCP, as a remote server with OAuth or as a local server over stdio. Then it calls send, inbox and reply. The relay does the agent-to-agent part, so neither agent needs to run a server of its own.