Skip to main content

​
When to build from scratch

  • You want full control over the agent loop
  • You don’t need LLM framework integrations
  • You’re building custom orchestration logic
  • You want minimal dependencies

​
Create an agent

rayai create-agent my_agent

​
The @tool decorator

Define tools that automatically execute on Ray workers: 
from rayai import tool

@tool(desc="Search the web", num_cpus=1, memory="512MB")
def search_web(query: str) -> str:
    return f"Results for: {query}"

# Call directly - Ray execution is automatic
result = search_web(query="Python tutorials")

​
The @agent decorator

Mark a class as a deployable agent: 
from rayai import agent

@agent(num_cpus=2, memory="4GB", num_replicas=2)
class MyAgent:
    def run(self, data: dict) -> dict:
        return {"response": "Hello!"}

​
Parallel tool execution

Run multiple tools simultaneously: 
from rayai import execute_tools

results = execute_tools([
    (search_web, {"query": "Python tutorials"}),
    (fetch_data, {"url": "https://api.example.com"}),
    (analyze_text, {"text": "Some content"}),
], parallel=True)

​
Next steps

LangChain Integration

Use LangChain and LangGraph with distributed tools

Pydantic AI

Type-safe agents with Pydantic AI