Skip to content

Introduction to Agent SDK

Welcome to the Agent SDK, a lightweight, zero-dependency Python framework designed for building intelligent LLM Agents and Multi-Agent Swarms.

Whether you are building a simple chatbot, a complex research assistant, or an autonomous swarm of agents working together, Agent SDK provides the primitives you need without the bloat.

Why Agent SDK?

  • Lightweight & Fast: No heavy dependencies. Pure Python.
  • Model Agnostic: Works with OpenAI, Gemini, Anthropic, Ollama, and more.
  • Middleware First: Easily add capabilities like Memory (RAG), Safety, and Logging via a robust middleware system.
  • Swarm Ready: Built-in support for multi-agent collaboration and handoffs (AgentSwarm).

Core Concepts

  • Agents: The brain of your application. An Agent wraps an LLM with specific instructions and tools.
  • Runners: The runtime environment that executes the Agent's reasoning loop.
  • Tools: Functions that Agents can call to interact with the real world (APIs, File System, etc.).
  • Middleware: Plugins that intercept and modify the Agent's behavior (e.g., adding long-term memory).

Getting Started

Ready to build?

  1. Installation: Set up your environment.
  2. Quick Start: Build your first agent in 5 minutes.
  3. Examples: Check out real-world usage patterns.

Community

  • Quick Start --- Get up and running with a simple agent in seconds. ➡ Go to Quick Start

  • Agents & Tools --- Dive deep into creating powerful agents with custom tools. ➡ Learn about Agents

  • Middleware --- Enhance your agents with RAG, Safety checks, and Logging. ➡ Explore Middleware

  • :material-hive: Swarm Architecture --- Connect multiple agents into a collaborative mesh network. ➡ Build a Swarm