Welcome to Arsenal

Arsenal is Clustr's lightweight, unified AI agent framework, designed to help you build intelligent applications effortlessly. With its powerful, unified interface, you can seamlessly interact with multiple AI models, execute complex skills.

🚀 Easy start

Quickly build AI agents with minimal code

🔧 Customize

Add custom skills to extend functionality

⚡ Scale

Efficient async design

🌐 Connect

Access 300+ LLMs through one unified API

What Makes Arsenal Different?

Lightweight Core: Only 12KB - no bloated dependencies
Provider Agnostic: Work with OpenAI, Anthropic, X-AI, OpenRouter, and more
Modern Design: Built with Python 3.12+, Pydantic, and async-first architecture
Developer-Focused: Clear documentation, type hints, and intuitive API

Quick Example

import os
from dotenv import load_dotenv
from arsenalpy.agents import Agent, AgentConfig
import asyncio

load_dotenv()

# Initialize an agent
agent = Agent(
    name="my_assistant",
    provider="openai",  # Also supports: anthropic, openrouter, and more
    system_prompt="You are a helpful AI assistant.",
    model="o3-mini",
    api_key=os.getenv("OPENAI_API_KEY"),
    config=AgentConfig(
        temperature=0.7,
        max_completion_tokens=1000,
    ),
)

# Use the agent
async def main():
    # Simple completion
    response = await agent.do("What are three creative uses for AI in Blockchain?")
    print(response)

    print("\n--- Streaming Response ---\n")

    # Streaming response
    async for chunk in agent.do_stream("Explain EVM in simple terms."):
        print(chunk, end="", flush=True)

if __name__ == "__main__":
    asyncio.run(main())