Open Protocol · Apache 2.0

The Coordination Layer
for Agentic AI

Multi-agent systems need infrastructure, not more messaging. SEMP enables O(n) coordination through shared environment signals that decay over time — inspired by 60 years of swarm intelligence research.

Read the Docs View on GitHub

Full-stack MVP built

IEI Bayreuth Finalist

Raising $3-5M Seed

The Challenge

Agent-to-agent messaging
doesn't scale

As you add agents, direct communication creates exponential complexity. We solved this with a fundamentally different architecture.

Today's Approach

Direct messaging: O(n²)

10 agents = 90 connections. 100 agents = 9,900. Every new agent multiplies the chaos.

SEMP Protocol

Shared environment: O(n)

Agents read and write signals to a shared layer. Adding agents is linear, not exponential.

Protocol

How SEMP Works

Agents coordinate through environmental signals — like pheromones in nature. No direct communication needed.

Agent emits signal

Write intent to the shared environment layer with type, scope, and strength

Environment processes

SEL stores, decays, amplifies, and makes signals queryable in real-time

Agents read & react

Other agents observe the environment state and make autonomous decisions

Five signal types cover all coordination patterns

PROPOSAL — "I suggest..."

READY — "I'm done"

TEST — "Did it work?"

REWARD — "That was good"

PENALTY — "That was bad"

S(t) = S₀ × λ^t — signals decay, good paths strengthen

Developer Experience

Integrate in minutes,
not months

SEMP is framework-agnostic. One API call to emit a signal, one to read the environment. Works with LangGraph, CrewAI, AutoGen, or your custom stack.

REST API with full TypeScript types and Zod validation
WebSocket for real-time signal updates
Full audit trail with trace IDs for every decision
Python & TypeScript SDKs with LangGraph connector

emit_signal.py

# Emit a coordination signal
import requests

response = requests.post(
    "https://selsemp.greenbee-ai.info/api/events",
    json={
        "type": "PROPOSAL",
        "scope": "logistics/warehouse-7",
        "pattern": "reroute-shipment",
        "strength": 0.85,
        "agentId": "planner-agent-01",
        "metadata": {
            "reason": "Delay detected",
            "priority": "high"
        }
    }
)

# Other agents see this signal,
# react autonomously, and the
# signal decays over time.

Traction

Built, tested, running

Not a whitepaper — a working system with real code, real tests, and a live demo.

13,930

Lines of Code

Full-stack TypeScript

Automated Tests

Vitest, all passing

Verticals

Farming + Robotics modules

Real-time

Dashboard

WebSocket observability

TypeScript React Express WebSocket PostgreSQL Drizzle ORM Vite Docker

Use Cases

One protocol, many
coordination domains

SEMP is domain-agnostic infrastructure. Any system with multiple autonomous agents benefits from shared environment coordination.

Featured Vertical

Market Context

Infrastructure layers
become industry standards

Every major computing paradigm needed a coordination layer. Agentic AI is next.

Confluent $10B+

Built Kafka into the standard for event streaming. Open protocol, enterprise platform.

Kafka is to streaming what SEMP is to agent coordination

HashiCorp $5B IPO

Terraform became the standard for infrastructure-as-code. Free CLI, paid platform.

Terraform is to infra what SEMP is to multi-agent systems

Datadog $35B+

Defined the monitoring category. Observability layer that became essential infrastructure.

Datadog is to monitoring what SEL is to agent observability

For Investors

We're raising $3-5M Seed to
build the coordination standard.

18-month runway to 3+ enterprise pilots, 1,000+ GitHub stars, and Series A readiness at $30-50M.

View Pitch Deck Download Pitch Deck

$3-5M

Seed Round

18 mo

Runway

$30-50M

Series A Target

$47B+

TAM by 2030

The Coordination Layer
for Agentic AI