AWS launches AgentCore system and agentic marketplace

Focus on Agentic AI

The vendor's focus on agents and agentic AI stems from AWS's conviction that agentic AI represents a fundamental shift that is transforming the tech landscape.

AWS is backing up its position on agentic AI by investing an additional $100 million in its AWS generative AI Innovation Center.

"The focus is on helping organizations truly get that hyper automation ability," said Taimur Rashid, managing director of generative AI innovation and delivery at AWS, during a media event on July 15. "Truly to take AI and not only generate and summarize but act on behalf of human beings as well."

Bradley Shimmin, an analyst with Futurum Group, said AgentCore is an extension of the AI frameworks and orchestration capabilities that AWS has presented to enterprises in the last couple of years.

"This is just a continuation of that and refinement of that and a consistent interface for developers to build teams that gives them access to some of the important tech AWS is shipping right now," Shimmin said.

AWS is creating an AI development toolchain that enables developers to build an agentic stack more easily, said Jason Andersen, an analyst with Moor Insights.

"It's geared toward the hardcore professional developer or enterprise developer versus what you're seeing from some of the other folks like Salesforce, for example, who are doing more low-code cell development," Andersen said.

However, he added that this is part of a concerted effort by the AI industry as a whole. Google offers much of what AWS offers with AgentCore and its agentic tools, but it's packaged differently. Microsoft offers similar tools, Andersen continued.

"What you're starting to see is more focus on AI developers using command line interface (CLI) tools," Andersen said. CLI tools are software applications used through a text-based interface to interact with a computer's OS. AWS joins Google, OpenAI and Anthropic in offering this capability.

"These types of stacks speed up the process of working with a CLI-based coding solution for agents," he said. "So, they kind of all go well together."

Mark Beccue, an analyst with Omdia, a division of Informa TechTarget, said targeting the developer alone might not be the right approach.

"They have disjointed messaging," he said. "When talking about agents, you must have the complete story."

He added that focusing on the developer might mean that AWS is not selling to those with the buying power.

An AWS partner

For John Balsavage, president of A&I Solutions, Inc., an IT consulting firm that works with AWS, AgentCore Observability is a key tool that could help with some AI agents' accuracy problems. The tool lets users trace, debug and monitor agents' performance in production environments.

While AWS and other AI vendors sometimes give themselves a passing mark if AI agents are 90% accurate, Balsavage told Informa TechTarget that is still not enough.

"Having observability [is key] to make sure we can drive to 100% accuracy," he said.

He added that another interesting tool AWS introduced is Kiro, an agentic IDE launched on July 14. Kiro works with any cloud provider or technology stack.

According to Balsavage, his company's developers have been trying to build a capability on AWS and are having trouble with prompting. Kiro could help solve that problem.

"It can be much more thorough at prompting," he said. "It can give you a better requirement."

While AgentCore and Kiro are needed capabilities, Shimmin said the introduction of AI Agents and Tools in the AWS Marketplace is also important. According to AWS, this is a centralized catalog of AI agents, agent tools and services where customers can build ready-to-integrate AI applications.

"In adding this … that is the fuel that will give AgentCore the ability to solve business problems," he said. "It's not just about the data model, it's about the business, the meaning of the data."

AWS also introduced Amazon S3 Vectors, a cloud object storage system with native vector support for AI technology.

Challenges

While AWS has long tried to be an ecosystem that provides users with as much choice as possible, it's now focusing on packaging its products to show users standard techniques or use cases, Andersen said. He added that the vendor is trying to be a medium for users between ‘hey this is too hard for me to figure out and too hard for me to put together,’” he said.

The challenge is whether this middle ground could truly help AWS or if it could become another way the tech giant will use to compete against Microsoft and GitHub, Andersen continued. He added that another challenge for AWS is being able to pull off the partnership ecosystem that it's starting to build.

"It's a little unfamiliar to them, so we'll see how they execute," he said.

AWS also revealed that TwelveLabs AI, a platform that turns video libraries into searchable, actionable assets, is now available on Amazon Bedrock.

AWS and Meta have also partnered to help startups build AI applications on Meta’s Llama open source large language model (LLM). Some 30 North American startups will have access to up to $200,000 in AWS credits and technical support from Meta and AWS experts.

Moreover, more than 6,600 AWS Academy students have free access to AWS Skill Builder's subscription-tier training content and AWS Certification exam vouchers.

Meanwhile, AWS-backed foundation model provider Anthropic launched an analytics dashboard for Claude Code. It enables enterprise admins to see how their teams use AI coding tools.

Esther Shittu is an Informa TechTarget news writer and podcast host covering AI software and systems.