Strands Agents SDK

Build AI agents using the model-driven Strands Agents SDK. Agents consist of three components: a model, tools, and a prompt.

Installation

pip install strands-agents strands-agents-tools --break-system-packages

Basic Agent

from strands import Agent

agent = Agent(system_prompt="You are a helpful assistant.")
response = agent("Hello, how can you help me?")
print(response)

Custom Tools

Use @tool decorator to convert Python functions into agent tools:

from strands import Agent, tool

@tool
def calculate_area(length: float, width: float) -> float:
    """Calculate rectangle area.
    
    Args:
        length: Rectangle length
        width: Rectangle width
    
    Returns:
        Area of the rectangle
    """
    return length * width

agent = Agent(tools=[calculate_area])
agent("What is the area of a 5x3 rectangle?")

Key requirements for tools:

• Clear docstring (used by LLM to understand when/how to use the tool)
• Type hints for all parameters and return value
• Descriptive parameter names

Class-Based Tools

For tools that share state or resources:

from strands import Agent, tool

class DatabaseTools:
    def __init__(self, connection_string: str):
        self.conn = self._connect(connection_string)
    
    def _connect(self, conn_str):
        return {"connected": True, "db": conn_str}
    
    @tool
    def query(self, sql: str) -> dict:
        """Execute SQL query.
        
        Args:
            sql: SQL query to execute
        """
        return {"results": f"Query: {sql}", "connection": self.conn}

db = DatabaseTools("postgres://...")
agent = Agent(tools=[db.query])

Model Providers

Default is Amazon Bedrock with Claude. Configure alternatives:

from strands import Agent
from strands.models import BedrockModel
from strands.models.ollama import OllamaModel
from strands.models.anthropic import AnthropicModel

# Amazon Bedrock (default)
agent = Agent(model=BedrockModel(
    model_id="us.anthropic.claude-sonnet-4-20250514-v1:0",
    region_name="us-west-2"
))

# Anthropic API (set ANTHROPIC_API_KEY env var)
agent = Agent(model=AnthropicModel(
    client_args={"api_key": "<KEY>"},
    model_id="claude-sonnet-4-20250514",
    max_tokens=1028
))

# Ollama (local)
agent = Agent(model=OllamaModel(
    host="http://localhost:11434",
    model_id="llama3.1"
))

MCP Integration

Connect to Model Context Protocol servers for external tools:

from strands import Agent
from strands.tools.mcp import MCPClient

# Stdio transport
mcp = MCPClient(transport="stdio", command="npx", args=["-y", "@modelcontextprotocol/server-filesystem"])

agent = Agent(tools=[mcp])
agent("List files in the current directory")

Multi-Agent Patterns

Agents as Tools

Wrap specialized agents as tools for an orchestrator:

from strands import Agent, tool

@tool
def research_assistant(query: str) -> str:
    """Research factual information.
    
    Args:
        query: Research question
    """
    researcher = Agent(system_prompt="You are a research expert.")
    return str(researcher(query))

@tool  
def code_assistant(task: str) -> str:
    """Write and explain code.
    
    Args:
        task: Coding task description
    """
    coder = Agent(system_prompt="You are a coding expert.")
    return str(coder(task))

orchestrator = Agent(
    system_prompt="Route tasks to the appropriate specialist.",
    tools=[research_assistant, code_assistant]
)

Swarm

Autonomous agent collaboration with shared context:

from strands import Agent
from strands.multiagent import Swarm

researcher = Agent(name="researcher", system_prompt="You research topics thoroughly.")
analyst = Agent(name="analyst", system_prompt="You analyze data and findings.")
writer = Agent(name="writer", system_prompt="You write clear reports.")

swarm = Swarm([researcher, analyst, writer])
result = swarm("Research AI trends and write a summary report")

Swarms enable emergent intelligence through:

• Shared working memory
• Autonomous handoffs between agents
• Dynamic task delegation

Graph

Deterministic workflows with defined execution order:

from strands import Agent
from strands.multiagent import GraphBuilder

researcher = Agent(name="researcher", system_prompt="Research the topic.")
reviewer = Agent(name="reviewer", system_prompt="Review and fact-check.")
writer = Agent(name="writer", system_prompt="Write the final output.")

builder = GraphBuilder()
builder.add_node(researcher, "research")
builder.add_node(reviewer, "review") 
builder.add_node(writer, "write")
builder.add_edge("research", "review")
builder.add_edge("review", "write")
builder.set_entry_point("research")

graph = builder.build()
result = graph("Write a report on quantum computing")

Conditional Edges

def needs_revision(state):
    return "needs revision" in state.results.get("review", {}).get("output", "").lower()

builder.add_edge("review", "research", condition=needs_revision)  # Loop back
builder.add_edge("review", "write", condition=lambda s: not needs_revision(s))

Streaming

Callback Handler

from strands import Agent
from strands.handlers import PrintingCallbackHandler

agent = Agent(callback_handler=PrintingCallbackHandler())
agent("Tell me a story")

Async Iterator

import asyncio
from strands import Agent

async def stream_response():
    agent = Agent()
    async for event in agent.stream_async("Tell me a story"):
        if hasattr(event, 'data'):
            print(event.data, end="", flush=True)

asyncio.run(stream_response())

Session Management

Persist conversations across sessions:

from strands import Agent
from strands.session import FileSessionManager

session_mgr = FileSessionManager(session_id="user-123", base_dir="./sessions")
agent = Agent(session_manager=session_mgr)

agent("Remember my name is Alice")
# Later session...
agent("What's my name?")  # Recalls "Alice"

Built-in Tools

The strands-agents-tools package provides:

• calculator - Math operations
• current_time - Get current time
• http_request - Make HTTP requests
• file_read, file_write, editor - File operations
• python_repl - Execute Python code
• shell - Run shell commands
• memory - Store/retrieve information
• retrieve - RAG retrieval
• mcp_client - Dynamic MCP server connections

from strands import Agent
from strands_tools import calculator, current_time, file_read

agent = Agent(tools=[calculator, current_time, file_read])

Pattern Selection Guide

Pattern	Use When
Single Agent	Simple tasks, direct tool use
Agents as Tools	Hierarchical delegation, clear specialist roles
Swarm	Exploration, brainstorming, emergent solutions
Graph	Deterministic workflows, conditional logic, loops

References

For detailed API documentation and advanced patterns, see:

• references/model-providers.md - Complete model provider configurations
• references/multi-agent-patterns.md - Advanced multi-agent architectures