A Python SDK for building autonomous AI agents and multi-agent orchestrations with native support for AWS Bedrock, local LLMs, and MCP tools.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install aws-strands
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install aws-strands using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Strands Agents SDK?
The Strands Agents SDK is a powerful Python library designed to streamline the creation and execution of autonomous AI agents. Originally developed by AWS, it provides a robust framework for building single-agent tasks or complex multi-agent workflows. It allows developers to integrate seamlessly with various model providers, including Amazon Bedrock, Anthropic, OpenAI, and local Ollama instances.
By leveraging Openclaw Skills like Strands, developers can implement advanced agentic patterns such as Swarms and Directed Acyclic Graphs (DAGs). The SDK is built with extensibility in mind, offering a rich set of built-in tools and the ability to connect to any Model Context Protocol (MCP) server, making it a versatile choice for modern AI application development.
Strands Agents SDK Use Cases
- Developing autonomous research assistants that summarize web content and write blog posts.
- Orchestrating multi-agent teams using the Swarm pattern for complex, handoff-based tasks.
- Building deterministic agent workflows with Graph-based execution for reliable production pipelines.
- Integrating local LLMs via Ollama for privacy-focused or cost-effective agentic processing.
- Creating stateful tools that interact with the local filesystem, shell, or external APIs.
How Strands Agents SDK Works
- Initialize an Agent by specifying a model provider such as BedrockModel or OllamaModel.
- Define custom tools using the @tool decorator or import from the strands-agents-tools library.
- Compose agentic logic using either a single agent, a Swarm for autonomous handoffs, or a Graph for structured dependencies.
- Manage state and history through built-in session managers supporting local files or Amazon S3.
- Execute prompts and handle tool-calling loops automatically via the SDK's core execution engine.
Strands Agents SDK Setup
To get started with these Openclaw Skills, install the SDK and tools using pip:
# Recommended: Install via pipx for CLI isolation
pipx install strands-agents-builder
# Alternatively, install directly into your environment
pip install strands-agents strands-agents-tools
Ensure you have your environment variables configured (e.g., AWS_ACCESS_KEY_ID, ANTHROPIC_API_KEY, or OPENAI_API_KEY) depending on your chosen model provider.
Strands Agents SDK Data Schema & Taxonomy
Strands organizes data through sessions and tool definitions. It uses a structured approach to manage the lifecycle of an agent's memory and capabilities.
| Entity | Storage Type | Description |
|---|---|---|
| Session State | JSON / S3 | Persists conversation history and tool context across runs. |
| Custom Tools | Python Scripts | Tools can be hot-reloaded from a directory if enabled. |
| Trace Data | OpenTelemetry | Captures lifecycle events, model invocations, and tool executions. |
| Config | Environment Vars | Handles API keys and regional settings for providers like Bedrock. |
Strands Agents SDK Advanced Features
- Multi-Agent Handoffs: Use the Swarm pattern to enable agents to autonomously hand off tasks to specialized peers.
- Model Context Protocol: Direct integration with MCP servers to expand agent capabilities without custom coding.
- Session Persistence: Native support for FileSessionManager and S3SessionManager to maintain long-running agent contexts.
- A2A Protocol: Standardized Agent-to-Agent communication for cross-network agent interoperability.
- Experimental Bidirectional Streaming: Real-time voice and text interaction with models like Nova Sonic and OpenAI Realtime.
SKILL.md
Loading
Related Openclaw Skills
rbw (Bitwarden CLI)
A high-performance CLI client for Bitwarden and Vaultwarden designed for secure credential retrieval and script integration.
Obsidian
A powerful automation skill for managing Obsidian vaults and Markdown notes via the command line using notesmd-cli.
Trinity Compress
A lightweight local script that compresses prompt files to minimize token usage in iterative AI development cycles.
Trinity Compress
A local script that compresses prompt files to minimize token costs in iterative AI development cycles.
Nexus Nova
A sophisticated personality architecture that transforms AI agents from simple scripts into autonomous entities with genuine identity and emotional depth.
Readwise MCP (mcporter)
Connect your Readwise and Reader library to AI agents using the Model Context Protocol for semantic search and document management.
Featured*
