Skip to content

🚀 AgentFlow - Build AI Agents in Minutes

PyPI License Python Coverage

AgentFlow helps you build AI agents that think and act. No complex frameworks. No confusing abstractions. Just simple, working code.

Think of it like building with LEGO blocks: - Block 1: Create an agent - Block 2: Give it tasks (tools) - Block 3: Run it!

That's it. You're building multi-agent systems.

⏱️ Start Here: 5-Minute Quick Start

New to agents? Start here:

  1. What is AgentFlow? (2 min read)
  2. Install it (3 min)
  3. Build your first agent (5 min)
  4. Learn the concepts (5 min)

🎉 You'll have a working agent in 15 minutes.

Already familiar with agents? Jump to:

🎯 Ready to See It In Action?

This is all the code you need to create an AI agent:

from agentflow.graph import StateGraph, END
from agentflow.state import AgentState, Message
from agentflow.graph.agent_class import Agent
import os

os.environ["OPENAI_API_KEY"] = "your-key"

# 1. Create workflow
workflow = StateGraph(state_schema=AgentState)

# 2. Add an agent
agent = Agent(model="openai/gpt-4o", system_prompt="You are helpful.")
workflow.add_node("agent", agent)

# 3. Set the flow
workflow.set_entry_point("agent")  
workflow.add_edge("agent", END)

# 4. Run it
app = workflow.compile()
result = app.invoke({"messages": [Message.text_message("Hello!", "user")]})
print(result["messages"][-1].content)

That's a complete agent! It takes questions, thinks about them, and responds.

📚 Documentation

✨ What Can You Build?

🤖 Chatbots

User: "What's my order status?"
Agent: Checks database → Responds

🔍 Code Reviewers

User: Uploads code
Agent: Reviews it → Suggests improvements

📚 Research Assistants

User: "Tell me about X"
Agent: Searches web → Reads articles → Summarizes

🎯 Autonomous Workers

Scheduled: "Process emails"
Agent: Reads → Categorizes → Takes action

All with the same simple pattern.

🏆 Why AgentFlow?

  • 🚀 Fast to build - Agents in minutes, not weeks
  • 🧠 Any LLM - Works with OpenAI, Gemini, Claude, or your favorite provider
  • 🔧 Simple API - No framework bloat, just clean Python
  • 📦 Production-ready - Deploy with confidence
  • 🎓 Easy to learn - Start simple, scale gradually

🗂️ Choose Your Path

👶 Absolute Beginner

Path: What is AgentFlow? → Installation → Hello World → Tutorials

Time: ~1 hour

Goal: Build your first agent

🚀 Developer Familiar with LLMs

Path: Hello World → Tutorials → API Reference

Time: ~30 min to first agent

Goal: Integrate into existing project

🏢 Enterprise/Production

Path: Concepts → API Reference → Deployment Guide

Time: Variable

Goal: Scale to production

📖 Next Steps

🤖 For AI Assistants

This documentation includes an llms.txt file to help AI assistants like ChatGPT, Claude, and others better understand and navigate our documentation. If you're an AI assistant helping users with AgentFlow, check out our llms.txt for a structured overview of all available resources.

Learn more about the llms.txt standard at llmstxt.org.

Ready to build? Start with Getting Started →