README.md

---
title: "AIPM - Agent Interoperability Protocol Models"
emoji: 🤝
colorFrom: blue
colorTo: purple
sdk: static
pinned: false
license: apache-2.0
tags:
  - ai-agents
  - multi-agent
  - interoperability
  - protocol
  - sdk
  - cryptography
  - python
---

# Agent Interoperability Protocol Models (AIPM)

## Vision

AIPM is the first open ecosystem for interoperable AI agents, enabling agents from different vendors (OpenAI, Claude, LangGraph, AutoGen, CrewAI, etc.) to communicate securely and consistently using the same protocol.

## Architecture Overview

### Core Protocol Components

1. **Identity Layer** - Agent identification and capabilities
2. **Capability Discovery** - Automatic discovery of skills and tools
3. **Secure Handshake** - TLS-inspired connection establishment
4. **Task Negotiation** - Accept/decline work based on capability
5. **Memory Exchange** - Efficient context sharing
6. **Trust Layer** - Reputation and reliability tracking
7. **Skill Marketplace** - Dynamic capability discovery
8. **Workflow Delegation** - Hierarchical task orchestration
9. **Economic Layer** - API billing and micropayments
10. **Standard Message Format** - JSON-based protocol

## Project Structure

```
aipm/
├── schemas/           # JSON schemas for protocol messages
├── sdk-python/        # Python reference SDK
├── sdk-javascript/    # JavaScript SDK (future)
├── sdk-rust/         # Rust SDK (future)
├── models/           # Fine-tuned AIPM models (future)
├── datasets/         # Training and benchmark datasets (future)
├── examples/         # Example implementations
└── docs/             # Protocol documentation
```

## Quick Start

### Installation

```bash
cd sdk-python
pip install -e .
```

### Basic Usage

```python
from aipm import AIPMAgent, AgentIdentity, Capabilities

# Create agent identity
identity = AgentIdentity(
    agent_id="my-agent-001",
    organization_id="my-org",
    name="My AI Agent",
    version="1.0.0",
    capabilities=Capabilities(
        skills=["text-generation", "code-review"],
        models=["gpt-4"],
        tools=["code-interpreter"],
    )
)

# Initialize agent
agent = AIPMAgent(identity)

# Initiate handshake with another agent
peer = AgentReference(
    agent_id="peer-agent-001",
    organization_id="peer-org"
)
hello_msg = agent.initiate_handshake(peer)
```

### Run Example

See `examples/basic_handshake.py` for a complete handshake between OpenAI and LangGraph agents:

```bash
cd examples
python basic_handshake.py
```

## Current Status

### ✅ Phase 1: COMPLETE

- [x] JSON schemas defined
- [x] Python SDK scaffolded
- [x] Identity & handshake models implemented
- [x] Basic agent implementation
- [x] Cryptographic foundation (Ed25519)
- [x] Example scripts

**See [PHASE1_COMPLETE.md](./PHASE1_COMPLETE.md) for full details**

### 🚧 Phase 2: In Planning

- [ ] Task negotiation framework
- [ ] Cryptographic message signing
- [ ] HTTP/WebSocket transport
- [ ] Enhanced error handling
- [ ] Comprehensive test suite

### 📋 Future Phases

**Phase 3: Advanced Features**
- [ ] Memory exchange protocol
- [ ] Trust scoring system
- [ ] Economic layer implementation

**Phase 4: Ecosystem**
- [ ] JavaScript SDK
- [ ] Rust SDK
- [ ] Fine-tuned AIPM models
- [ ] Benchmark datasets
- [ ] Public registry/marketplace

## Handshake Protocol

```
Agent A                          Agent B
   |                                |
   |------- HELLO ----------------->|
   |<--- CAPABILITY_EXCHANGE -------|
   |------- AUTHENTICATION -------->|
   |<--- PUBLIC_KEY_EXCHANGE -------|
   |------- TRUST_VERIFICATION ---->|
   |<--- READY ---------------------|
   |                                |
   [Ready for task delegation]
```

## Key Features

### Identity Layer
- Unique agent IDs
- Organization affiliations
- Capability declarations
- Trust scores
- Public key cryptography

### Secure Communication
- Ed25519 signatures
- Message authentication
- Session management
- Trust verification

### Interoperability
- Vendor-agnostic protocol
- Standardized message format
- Capability-based routing
- Cross-framework communication

## Use Cases

1. **Multi-Agent Workflows** - Agents from different vendors collaborate on complex tasks
2. **Skill Marketplace** - Discover and delegate to specialized agents
3. **Trust Networks** - Build reputation across agent interactions
4. **Economic Coordination** - Fair billing and micropayments between agents
5. **Memory Sharing** - Efficient context exchange without duplication

## Technical Stack

- **Protocol:** JSON-based message format
- **Cryptography:** Ed25519 (EdDSA)
- **Python SDK:** Pydantic, cryptography, httpx
- **Schemas:** JSON Schema Draft 2020-12

## Documentation

- [Phase 1 Complete](./PHASE1_COMPLETE.md) - Detailed Phase 1 documentation
- [SDK README](./sdk-python/README.md) - Python SDK documentation
- [Schemas](./schemas/) - JSON schema specifications

## Examples

- [basic_handshake.py](./examples/basic_handshake.py) - Complete handshake demo
- [verify_phase1.py](./examples/verify_phase1.py) - Verification script

## Contributing

We welcome contributions! Areas of focus:

- Protocol design and specification
- SDK implementations (Python, JS, Rust, Go)
- Example applications
- Documentation and tutorials
- Test coverage
- Benchmark datasets

## Roadmap

**Q3 2026**
- ✅ Phase 1: Core protocol and SDK
- 🚧 Phase 2: Task negotiation and transport

**Q4 2026**
- Phase 3: Advanced features (memory, trust, economic)
- Additional language SDKs

**2027**
- Fine-tuned AIPM models
- Public agent registry
- Enterprise features
- Ecosystem growth

## License

Apache 2.0

## Contact

- GitHub: https://github.com/aipm/aipm
- Documentation: https://docs.aipm.org
- Community: https://discord.gg/aipm

---

**Building the future of interoperable AI agents** 🚀