Skip to main content
Cloud / AWS / Products / Strands Agents: Open Source SDK for AI Agents

Strands Agents: Open Source SDK for AI Agents

Strands Agents is an open source toolkit from AWS for production AI agents, available for Python and TypeScript under the Apache License 2.0.

Developer Tools
Pricing Model Open source under the Apache License 2.0, no licence fees; costs arise for the models used and the runtime environment
Availability Usable as an open source SDK for Python and TypeScript; runs in the environment of your choice
Data Sovereignty Depends on the chosen model provider and runtime environment
Reliability No SLA, as this is an open source SDK; the SLAs of the AWS services used apply SLA

What is Strands Agents?

Strands Agents is an open source toolkit for building production AI agents. It is available as an SDK for Python and TypeScript under the Apache License 2.0 and, per the project site, was built from production systems inside Amazon. AWS introduced the project on 16 May 2025.

The approach is model-driven: instead of modelling the sequence as a workflow in advance, the language model plans and directs its own steps and calls the tools it has been given. At its core an agent consists of a model, tools, and a prompt.

Core Features

  • Agent loop: Agents execute iteratively, making decisions and calling tools until the task is complete
  • Tools: Defined via the @tool decorator in Python or via schema objects in TypeScript
  • MCP support: Model Context Protocol clients are integrated, so published MCP servers can be used as tools
  • Hooks: Event-based interception points such as BeforeToolCallEvent and AfterToolCallEvent for monitoring, validating, and modifying agent behaviour
  • Conversation management: Built-in handling of conversation context, including summarisation and sliding-window approaches
  • Multi-agent patterns: Built-in patterns such as Agent-as-Tool and Swarm for coordinating several agents
  • Model independence: Amazon Bedrock, Anthropic Claude API, Llama API, Ollama, OpenAI via LiteLLM, and custom providers
  • Observability: Telemetry data emitted through OpenTelemetry to any OTEL-compatible backend

Typical Use Cases

Agents in production: The toolkit explicitly targets production agents and ships with safety, observability, and control features.

Connecting existing systems: Through MCP servers and custom tools, agents reach existing APIs and knowledge sources.

Multi-agent scenarios: Patterns such as Agent-as-Tool and Swarm coordinate several agents on one task.

Local development, cloud operation later: Ollama for local development, then deployment on AgentCore, Lambda, Fargate, EKS, Docker, or via Terraform.

Benefits

  • Open source under the Apache License 2.0, no licence fees
  • SDKs for Python and TypeScript
  • Model- and cloud-independent; application code stays the same when you switch
  • Standardised tool integration through MCP
  • Observability through OpenTelemetry instead of proprietary telemetry
  • Foundation of several AWS agent products and therefore close to their runtime environments

Integration with innFactory

As an AWS Reseller, innFactory supports you with Strands Agents: selecting model providers, designing tool and MCP integration, establishing guardrails and approval steps, setting up observability through OpenTelemetry, and deploying on Amazon Bedrock AgentCore, AWS Lambda, AWS Fargate, or Amazon EKS.

Typical Use Cases

AI agents in production
Tool calls via MCP
Multi-agent systems
Observability for agents

Technical Specifications

Languages Python and TypeScript
License Apache License 2.0
Observability OpenTelemetry (OTEL)

Frequently Asked Questions

What is Strands Agents?

Strands Agents is described as the open source toolkit for building production agents. It is available as an SDK for Python and TypeScript and, per the project site, was built from production systems inside Amazon. AWS introduced the project on 16 May 2025.

Under which licence is Strands Agents released?

Strands Agents is released under the Apache License 2.0.

What does the model-driven approach mean?

The model-driven approach leverages the capabilities of modern language models to plan, reason, and execute tasks autonomously. Instead of defining complex workflows up front, the model dynamically directs its own steps and uses tools in order to accomplish the specified task. The three core components are model, tools, and prompt.

Which model providers are supported?

The documentation names Amazon Bedrock, the Anthropic Claude API, Meta's Llama API, Ollama for local development, OpenAI via LiteLLM, and custom provider implementations. The project site emphasises model independence: your code stays the same across any model and any cloud.

How are tools connected?

Tools are defined via the @tool decorator in Python and via schema objects in TypeScript. Strands additionally integrates Model Context Protocol (MCP) clients, so published MCP servers can be used as tools.

Which deployment and observability options exist?

The project site names AgentCore, Lambda, Fargate, EKS, Docker, and Terraform as deployment targets. For observability, Strands uses OpenTelemetry (OTEL) to emit telemetry data to any OTEL-compatible backend. Hooks such as BeforeToolCallEvent and AfterToolCallEvent allow monitoring, validation, and modification of agent behaviour.

Note: All product information on this page has been compiled with care, but is provided without guarantee and may be outdated or incomplete. Cloud services evolve rapidly — features, pricing, SLAs, and availability change frequently. Authoritative and up-to-date information can only be found on the official product page of AWS (official documentation). This page does not represent an offer by AWS.

Quick Links

AWS Cloud Expertise

innFactory is an AWS Reseller with certified cloud architects. We provide consulting, implementation, and managed services for AWS.

Ready to start with Strands Agents: Open Source SDK for AI Agents?

Our certified AWS experts help you with architecture, integration, and optimization.

Schedule Consultation