https://github.com/josharsh/aa-kit

AA Kit is a Python framework designed to build AI agents that naturally compose into ecosystems. Every agent is simultaneously a standalone agent, an MCP server, and an MCP client - creating true interoperability across the entire AI landscape.
https://github.com/josharsh/aa-kit

Last synced: 4 months ago
JSON representation

AA Kit is a Python framework designed to build AI agents that naturally compose into ecosystems. Every agent is simultaneously a standalone agent, an MCP server, and an MCP client - creating true interoperability across the entire AI landscape.

• Host: GitHub
• URL: https://github.com/josharsh/aa-kit
• Owner: josharsh
• Created: 2025-07-08T07:30:04.000Z (about 1 year ago)
• Default Branch: main
• Last Pushed: 2025-07-08T07:35:47.000Z (about 1 year ago)
• Last Synced: 2025-12-16T18:21:45.288Z (7 months ago)
• Size: 1.95 KB
• Stars: 0
• Watchers: 0
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# 🤖 AA Kit: Agent As Everything



  

  

  





  
🚀 One Agent, Many Forms

  
Your agent is a tool. Your agent is a server. Your agent is a function.


  With AA Kit, agents transcend categories.



---

## 🌟 The Philosophy

> **"When you realize an agent can be a tool for another agent, and that tool is also an agent, you understand: agents are the universal primitive of AI."**

AA Kit isn't just another agent framework—it's a paradigm shift. We believe agents shouldn't be boxed into roles like "chatbot" or "assistant." Instead, agents are the **universal primitive** of AI systems, capable of being:

- 🔧 **Tools** for other agents

- 🌐 **Servers** via MCP protocol  

- ⚡ **Functions** that compose naturally

- 🏗️ **Building blocks** for AI ecosystems

## ✨ What Makes AA Kit Different

### 🎯 **Agent = Function**

```python

# Three parameters. Returns intelligence. No complexity.

agent = Agent(

    name="assistant",

    instruction="You are a helpful assistant", 

    model="gpt-4"

)

```

### 🌐 **Agent = Server**

```python

# Every agent exposes itself as an MCP server instantly

agent.serve_mcp()

# Now available at: http://localhost:8080/mcp

```

### 🔗 **Agent = Tool**

```python

# Pass one agent to another as a tool

writer = Agent(

    name="writer",

    instruction="You write articles",

    model="gpt-4",

    tools=[researcher_agent]  # Natural composition

)

```

### 🚀 **Agent = Production**

Rate limits, retries, caching, circuit breakers—not added later, but built into the constructor from day one.

## 🎨 Core Features

### 🏗️ **Universal Composition**

Agents compose naturally because they're all the same primitive underneath. No special interfaces or adapters needed.

### 🌐 **MCP Native**

Every agent is automatically MCP-compatible. Zero configuration, instant interoperability with the entire MCP ecosystem.

### ⚡ **Production Ready**

Built-in rate limiting, retry logic, circuit breakers, and caching. Scale from prototype to production without refactoring.

## 📊 By The Numbers



| **3** | **0** | **∞** | **1** |

|-------|-------|-------|-------|

| Lines to Production | Config for MCP | Composition Depth | Universal Protocol |



## 🔮 Coming Soon

The open source release of AA Kit is being finalized. Here's what you can expect:

### 📦 **Core Package**

- Pure Python implementation

- Zero external dependencies for core functionality

- Support for OpenAI, Anthropic, and local models

- Built-in memory backends (Redis, SQLite, PostgreSQL)

### 🛠️ **Developer Experience**

- Rich CLI with interactive agent builder

- Hot-reloading for development

- Comprehensive debugging tools

- Extensive examples and tutorials

### 🌐 **MCP Ecosystem**

- Automatic server generation

- Client connection helpers

- Protocol validation and testing

- Community tool registry

### 📚 **Documentation**

- Comprehensive API reference

- Architecture deep-dives

- Production deployment guides

- Framework comparison analysis

## 🎯 Quick Preview

While we prepare the full release, here's a taste of the AA Kit experience:

```python

from aakit import Agent

# Create a research agent

researcher = Agent(

    name="researcher",

    instruction="You research topics thoroughly and provide detailed analysis",

    model="gpt-4"

)

# Create a writer that uses the researcher

writer = Agent(

    name="writer", 

    instruction="You write engaging articles based on research",

    model="claude-3",

    tools=[researcher]  # Composition magic

)

# Serve both as MCP servers

researcher.serve_mcp(port=8080)

writer.serve_mcp(port=8081)

# Now available to the entire MCP ecosystem!

```

## 🚀 Get Notified

Want to be among the first to build with AA Kit?

- ⭐ **Star this repo** to get notified when we launch

- 🐦 **Follow us** on [Twitter](https://twitter.com/aakit_ai) for updates

- 📧 **Join our mailing list** at [aakit.dev](https://aakit.dev)

- 💬 **Join our Discord** for early access discussions

## 🤝 Early Access

Interested in contributing or getting early access? We're looking for:

- 🏗️ **Framework developers** familiar with agent architectures

- 🧪 **Beta testers** building production AI systems  

- 📝 **Technical writers** passionate about developer experience

- 🎨 **UI/UX designers** for our developer tools

Reach out at [hello@aakit.dev](mailto:hello@aakit.dev)

## 🌟 The Vision

AA Kit represents a fundamental shift in how we think about AI agents:

> **Not as isolated applications, but as composable primitives that form the building blocks of intelligent systems.**

Every agent can be:

- A standalone application

- A tool for other agents  

- A server in a distributed system

- A function in a larger computation

This isn't just code—it's a new way of architecting AI.

## 📄 License

MIT License - see [LICENSE](LICENSE) for details.

---



  🤖 Agent As Everything. Everything As Agent. 🤖

  



  The future of AI is composable.