AWS open-sources Strands Agents SDK to ease AI agent development – News

Amazon Web Services Inc. today released Strands Agents SDK, an open-source toolkit for developing artificial intelligence agents.

The project’s code is available on GitHub. 

AWS originally created Strands Agents SDK to support the work of its internal development teams. There were already tools on the market for building AI agents, but the cloud giant found that can require a significant amount of time to configure. “Where it used to take months for Q Developer teams to go from prototype to production with a new agent, we’re now able to ship new agents in days and weeks with Strands,” AWS Senior Principal Engineer Clare Liguori detailed in a blog post.

Strands Agents SDK allows developers to power their agents with large language models from multiple providers. It works with Amazon Bedrock, an AWS service that offers hosted foundation models, and Anthropic PBC’s application programming interface. The API provides access to the company’s cloud-based LLMs.

There’s also support for two open-source tools called Ollama and LiteLLM. The former technology allows developers to run open-source LLMs on their local machines. LiteLLM, in turn, provides a single interface for interacting with more than 100 cloud-based LLM APIs. That removes the need for software teams to familiarize themselves with each API.

After developers specify the LLM that they wish to turn into an agent, they can use Strands Agents SDK to integrate it with third-party applications. The toolkit provides access to thousands of MCP clients. Those are programs that enable an LLM to retrieve data from an external system or perform actions in that system.

Strands Agents SDK also includes more than 20 prebuilt tools that AI agents can use to complete tasks. One of those tools, Retrieve, is powered by AWS’ Amazon Bedrock Knowledge Bases service. It allows AI agents to automatically find tools that they can use to process prompts.

Another pre-built tool allows developers to create applications that use not one but multiple AI agents to perform tasks. AWS plans to enhance the capability in the future adding support for the Agent2Agent, or A2A, protocol. It’s an open-source technology that allows agents to coordinate their work by exchanging data with one another.

“Strands can orchestrate across multiple agents in a variety of multi-agent collaboration patterns,” Liguori wrote. “By modeling sub-agents and multi-agent collaboration as tools, the model-driven approach enables the model to reason about if and when a task requires a defined workflow, graph, or swarm of sub-agents.”

The toolkit also manages the data that AI agents process. It can store user prompts, an AI agent’s responses to those prompts and output from the applications that the agent relies on to complete tasks. Strands Agents SDK can filter unnecessary data to reduce infrastructure costs. 

“Multiple teams at AWS already use Strands for their AI agents in production, including Amazon Q Developer, AWS Glue, and Amazon VPC Reachability Analyzer,” Liguori detailed. 

Image: AWS