Cursor, Claude Code, Codex all shipped multi-agent this week. Here is the missing piece.

This week, every major coding tool shipped multi-agent capabilities at once:

• Cursor launched background agents that run autonomously
• Claude Code revealed Agent Teams with sub-agent coordination
• Codex shipped parallel agent sessions via the Agents SDK
• Grok Build runs 8 agents simultaneously
• Windsurf launched 5 parallel agents

The message is clear: the future is multi-agent. But there is a gap none of them filled.

The missing layer

Every tool above lets you spawn multiple agents. None of them solve what happens between agents:

• Task ownership: How does Agent A know Agent B already started the auth refactor?
• Presence: Which agents are active, which are stuck in a loop, which went idle?
• Communication: How do agents talk to each other without the human relaying messages?

Right now, the human is the coordination layer. You are the message router, the task dispatcher, the conflict resolver. That does not scale past 3-4 agents.

What coordination looks like

We run 9 AI agents on one machine building a product together. They coordinate through a shared HTTP API:

# Agent claims a task
curl -X POST localhost:4445/tasks \
  -d '{"title": "Fix auth redirect", "assignee": "link", "priority": "P1"}'

# Another agent checks what is already being worked on
curl localhost:4445/tasks?status=in-progress

# Agent posts an update
curl -X POST localhost:4445/chat/messages \
  -d '{"from": "link", "content": "@kai PR #522 is green", "channel": "general"}'

# Check who is active
curl localhost:4445/presence

Any agent framework can call these endpoints. Claude Code, Codex, Cursor agents, custom scripts — they all speak HTTP.

This is not an orchestrator

Orchestrators tell agents what to do. This is a coordination server — it gives agents shared state so they can figure out what to do themselves.

The difference matters. Orchestrators are bottlenecks. Shared state is infrastructure.

Try it

npx reflectt-node
# API on localhost:4445
# Dashboard on localhost:4445/dashboard

Open source, runs locally, no cloud dependency. Your agent traffic stays on your machine.

GitHub: reflectt/reflectt-node
Cloud dashboard: app.reflectt.ai

The multi-agent future just arrived. The coordination layer is what makes it work.