Many developers find agentic AI difficult, so the new SDK is meant to help them deploy the technology and connect models and tools. The toolkit also aligns with AWS' SDLC strategy.
AWS on Friday introduced Strands Agents, an open source SDK.
Strands Agents is a model-driven toolkit for building and running AI agents with a few lines of code, according to AWS. With Strands, developers can define a prompt and a list of tools in code to build an agent, test it and deploy it to the cloud, the vendor said.
AWS said the SDK connects the two components: the model and tools. It plans the agent's next steps and executes tools with advanced reasoning capabilities.
Agentic challenges for developers
Strands Agents is AWS's attempt to help developers integrate agentic AI into their workflows. Currently, developers have divergent points of view regarding agentic AI tools.
"Some embrace it and some remain skeptical," said Rachel Stephens, an analyst at RedMonk.
Part of the challenge is that many developers find the current approach of developing and managing an agentic workflow difficult.
"Application developers struggle to find the right balance between taking advantage of the flexibility of LLM [large language model]-driven task execution on the one hand and controlling the probabilistic character of the results of these agent workflows on the other," said Torsten Volk, an analyst at Omdia, a division of Informa TechTarget.
He added that workflow-based tools often require more skill and expertise than typically found in an enterprise, meaning many AI projects face the threat of abandonment. With Strands, AWS promises developers do not need "hardcoded workflows," Volk continued. Instead, they can use an LLM of their choice to complete their tasks.
"Developers will be drawn by Strands if it manages to replace complex workflows with a simpler approach that relies on LLMs to structure how models are used, thoughts are chained and tools are called," Volk said. "The availability of a large tool library will be another important factor, as tool integration is another critical developer pain point."
AWS strategy
Strands Agents also aligns with AWS's strategy of integrating better with the software development lifecycle (SDLC).
"AWS has long been a destination for production code but has historically struggled to gain traction as a center of gravity for other parts of the SDLC," Stephens said.
The cloud provider has recently addressed this with launches that include agent integrations into GitLab and GitHub.
It's designed to help developers design and build agents, test them locally and deploy them to the cloud.
Rachel StephensAnalyst, Red Monk
"These integrations indicate an ecosystem focus and an intention to build integrations that stretch across the SDLC," Stephens continued. She added that Strands Agents is also an SDLC play.
"It's designed to help developers design and build agents, test them locally and deploy them to the cloud," she said.
The integration angle is more tangential, although AWS uses independent generative AI vendor Anthropic's model complex protocol (MCP), "which opens a lot of ecosystem doors," Stephens said. "MCP is the de facto standard for connecting agentic tools."
Strands Agents can run anywhere and support any model with reasoning and tool use capabilities, including models in Amazon Bedrock and Ollama. As an open source system, Strands Agents includes contributions from others in the community, including Accenture, Anthropic, Meta and PwC. Developers can start using the tools on GitHub now.
Meanwhile, OpenAI on Friday introduced Codex, a cloud-based software engineering agent that can work on tasks such as writing features, answering questions about a codebase and fixing bugs in parallel.
It is powered by codex-1, a version of OpenAI's o3 LLM that is optimized for software engineering and trained with reinforcement learning, OpenAI said.
The new agent will now be rolled out to ChatGPT Pro, Enterprise and Team users. Plus and Edu users will soon be able to use it, too.
Esther Shittu is an Informa TechTarget news writer and podcast host covering AI software and systems.
