Skip to main content

Quick start

Get up and running in 5 minutes.

Installation​

pip install pureagents
export MISTRAL_API_KEY=your-key

Your first agent​

from pure_agents import Agent, tool


@tool
def greet(name: str) -> str:
    """Greet someone by name."""
    return f"Hello, {name}!"


agent = Agent(tools=[greet])
result = await agent.run("Say hello to Maria")

Common patterns​

Streaming​

async for event in agent.stream("Tell me a story"):
    if event.type == "text":
        print(event.content, end="", flush=True)

Structured outputs​

from dataclasses import dataclass


@dataclass
class Analysis:
    sentiment: str
    confidence: float


result = await agent.run("Analyse: I love this!", output=Analysis)

Memory​

agent = Agent(session="my-chat")

Different providers​

agent = Agent(provider="openai")
agent = Agent(provider="anthropic")
agent = Agent(provider="openai", fallback="mistral")  # Failover

Templates​

agent = Agent(template="coder")
agent = Agent(template="researcher")

Reliability​

agent = Agent(
    retries=3,
    timeout=30.0,
    max_messages=20,
    cache=True,
)

Tool control​

@tool(timeout=10, group="web")
def search(query: str) -> str:
    """Search with timeout."""
    return "..."


agent = Agent(
    tools=[search],
    tool_choice="required",
    enabled_groups=["web"],
)

Batch processing​

results = await agent.batch(["Q1", "Q2", "Q3"])

Images​

result = await agent.run(
    "Describe this image",
    images=["photo.jpg"],
)

Validation​

def is_json(response: str) -> bool:
    try:
        json.loads(response)
        return True
    except:
        return False


agent = Agent(validator=is_json, validation_retries=2)

Usage tracking​

await agent.run("Hello")
print(agent.usage.total_tokens)
print(agent.usage.cost("openai"))

Sync usage​

result = agent.run_sync("Hello")

Next steps​