strands-agents/sdk-python

★ 5,843PythonAudience · developerComplexity · 3/5Setup · moderate

Mindmap

mindmap
  root((Strands Agents SDK))
    What it does
      Build AI agents in Python
      Call tools automatically
      Multi-agent workflows
      Streaming responses
    Model Providers
      Amazon Bedrock
      Anthropic Claude
      OpenAI
      Ollama local models
    Key Features
      MCP protocol support
      Hot-reload tool files
      Swap models easily
    Use Cases
      Research agents
      Customer support bots
      Workflow automation

mindmap root((Strands Agents SDK)) What it does Build AI agents in Python Call tools automatically Multi-agent workflows Streaming responses Model Providers Amazon Bedrock Anthropic Claude OpenAI Ollama local models Key Features MCP protocol support Hot-reload tool files Swap models easily Use Cases Research agents Customer support bots Workflow automation

Click or tap to explore — scroll the page freely

Things people build with this

USE CASE 1

Build an AI agent that reads files and browses the web to research a topic and write a summary

USE CASE 2

Create a customer support bot that calls your own backend tools to look up orders or open tickets

USE CASE 3

Run multiple specialized agents in parallel, each solving a different part of a complex task

USE CASE 4

Swap between Claude, GPT-4, and a local Ollama model with just a config change to compare results

Tech stack

PythonAmazon BedrockAnthropic ClaudeOpenAIOllamaModel Context Protocol

Getting it running

Difficulty · moderate Time to first run · 30min

Default provider is Amazon Bedrock and requires AWS credentials, other providers need their own API keys.

No license information was mentioned in the explanation.

In plain English

Strands Agents is a Python library for building AI agents: programs that receive a goal written in plain language and then take a series of steps to reach it, calling tools and making decisions along the way. The library is designed so that a working agent can be created in just a few lines of code, with more advanced setups available as needs grow. The SDK connects to many different AI model providers, meaning the underlying intelligence driving the agent can come from Amazon Bedrock, Anthropic's Claude, Google Gemini, OpenAI, or locally-run models like Ollama and llama.cpp. You choose which one to use when setting up an agent, swapping to a different provider requires only a small configuration change. Agents in this framework can call tools, which are functions that give the AI the ability to take real actions: running calculations, reading files, browsing the web, or anything else a developer defines. There is also built-in support for the Model Context Protocol, an open standard that lets agents connect to a wide ecosystem of pre-built tool servers without custom integration work. More advanced scenarios are also supported. Multiple agents can work together, each handling a different part of a larger task. Streaming is available so that responses appear incrementally rather than all at once, which matters for conversational or voice-driven applications. The library can also watch a local folder for new tool files and load them automatically while a program is running. Installation is through pip, Python's standard package manager. The default model provider is Amazon Bedrock, which requires AWS credentials, but the documentation covers setup for all other supported providers as well.

Copy-paste prompts

Prompt 1

Using the Strands Agents Python SDK, write a minimal agent that uses Claude as the model and has a single tool that reads a local text file and returns its contents. Show the complete working code.

Prompt 2

I want to build a multi-agent workflow with Strands where one agent plans a task and a second agent executes each step. Show me how to wire two agents together so the planner hands off to the executor.

Prompt 3

I'm using Strands Agents with Amazon Bedrock but want to switch to OpenAI GPT-4. What is the exact config change I need to make and what credentials does OpenAI require?

Prompt 4

How do I add a custom tool to a Strands agent that calls my own REST API? Show a Python example where the tool makes an HTTP GET request and returns the JSON response to the agent.

Open on GitHub → Explain another repo

← strands-agents on gitmyhub — every repo by this author, as a profile.

Verify against the repo before relying on details.