flowing

Agent Interop Layer

An open interoperability layer for autonomous AI agents.

---

The Problem

AI agents are exploding in number.

Every week, new agents are built using:

• LangChain
• CrewAI
• AutoGPT-style systems
• Custom in-house stacks

But all of them run in isolation.

There is no common way for agents to:

• Discover each other
• Communicate
• Delegate tasks
• Share execution context
• Coordinate safely across frameworks

We are rebuilding the same glue logic over and over.

---

What This Project Is

This repository defines a minimal, framework-agnostic interoperability layer for AI agents.

It focuses on:

• A shared agent interface
• A neutral message & task exchange format
• Execution boundaries and permissions
• Simple coordination primitives (request, delegate, respond)

The goal is not to replace existing agent frameworks.
The goal is to let them talk to each other.

---

What This Project Is NOT

This is NOT:

• Another agent framework
• A model provider
• A hosted SaaS
• A closed ecosystem

This project is intentionally:

• Open
• Minimal
• Composable
• Model-agnostic

---

Why Now

• Models are becoming commodities.
• Agents are no longer demos — they are being deployed in real workflows.
• The bottleneck is no longer intelligence.
  The bottleneck is coordination.
• The missing piece is an execution and communication standard.

---

Core Ideas (Early Draft)

• Agents should be treated as networked actors
• Interoperability should not depend on a single vendor
• Control planes should be separate from models
• Coordination > raw intelligence

This repo is the place to experiment with those ideas.

---

Current State

Early stage / experimental

Right now, this repo contains:

• Initial interface definitions (BaseAgent)
• Draft protocol concepts
• Reference implementations (agent_server.py, my_agent_server.py)
• Examples of agents interacting

Everything is intentionally lightweight.

---

Quickstart (Run Your First Agent in 5 Minutes)

1. Clone the repository

git clone cd

2. Run the Reference Agent

python agent_server.py This starts AgentA on http://localhost:5000

3. Run the Reference Agent

python my_agent_server.py This starts AgentX on http://localhost:5001. Each dev can change the agent name and port.

4. Send Messages or Tasks Between Agents

From another terminal: curl -X POST http://localhost:5001/receive_message
-H "Content-Type: application/json"
-d '{"message": "Hello from AgentA"}'

curl -X POST http://localhost:5001/run_task
-H "Content-Type: application/json"
-d '{"description": "Test task", "price": 10}

Expected Output

• AgentX prints received messages
• AgentX executes tasks and prints balance updates
• Everything works without any private repo dependencies

---

Who This Is For

This project is for:

• Developers building autonomous agents
• Researchers exploring multi-agent systems
• Engineers tired of re-implementing glue code
• Anyone interested in agent coordination & orchestration

---

How to Get Involved

You can contribute by:

• Opening issues with real-world agent problems
• Proposing interface changes
• Implementing adapters for existing frameworks
• Stress-testing coordination patterns

If you're building agents today, your input matters.

---

Long-Term Vision

Become the interoperability layer that autonomous agents rely on — the infrastructure others build on top of.

• Not a product
• A standard