MACP: the Multi-Agent Cognition Protocol is a protocol for coordination between AI agents: messaging, handoffs, delivery semantics, findings, memory, goals, tasks, and durable context between independent agents while work is happening. Interested in feedback on the protocol surface, likely integration points, and which product categories would get the most value from this. https://www.macp.dev https://github.com/multiagentcognition/macp

Lightfield — AI-native CRM that builds itself and does work for you

AI-native CRM that builds itself and does work for you

I built MACP because once multiple autonomous AI agents are working in parallel, orchestration alone is not enough. Agents also need a shared way to signal ownership, share findings, surface important updates, and adapt while work is still in progress. MACP is a protocol for that coordination layer. It uses a lightweight shared message bus with durable delivery and priority-ranked signals, so agents can stay aligned without heavy infrastructure. You want specialized real-time agents preparing data, exploring different parts of the problem, and communicating only the most useful signals back to the agents that need them. MACP is designed to help make that possible: many cleaner contexts, connected by structured communication, instead of one giant context blob. I’d especially love feedback from people building multi-agent systems, coding agents, and orchestration tools: where does coordination break down today, and what would make something like this genuinely useful in production? Happy to answer questions about the protocol, design decisions, and use cases.

Lightfield — AI-native CRM that builds itself and does work for you

AI-native CRM that builds itself and does work for you

MACP: the Multi-Agent Cognition Protocol

A real-time cognition protocol for AI agents

A real-time cognition protocol for AI agents