AWS on Thursday revealed that it is rebuilding OpenSearch Serverless with architectural changes that support the new direction of search in the age of AI agents.
OpenSearch Serverless is Amazon’s service for deploying OpenSearch Service, a distributed search and analytics suite that provides fast access to large volumes of data.
The rebuild shows how enterprise search is becoming a foundational infrastructure for AI-driven operations, driving AI applications such as code generation, information retrieval and agentic systems. With this rebuild and a new open skills tool, AWS said it is providing enterprises with the infrastructure to adapt to that change.
“The problem with search was it was never anybody’s job to make it work, unless you were in the business of it, like Amazon.com is,” said Nick Patience, an analyst at Futurum Group. “If you were not that and you were just any other enterprise, an insurance company or bank, whatever, it really was not that deemed that important. [AWS is] saying that it is changing, and I would agree with that.”
OpenSearch Serverless and Agent Skills
AWS’ rebuild of OpenSearch Serverless supports the shift in search by making key improvements. According to AWS, OpenSearch has now become a search infrastructure for agentic AI. It has instant autoscaling, ready the moment an agent starts a task. Instant autoscaling is a compute capability that enables agents to respond to a request in seconds instead of minutes. Enterprises only pay for what they use, so whenever the agent is idle, there are no costs.
Also, a native AI platform integration is available on Vercel, a cloud platform for developers that developers use to build, deploy and host websites and applications, and as a Kiro Power — a downloadable plugin for the Kiro environment that provides AI agents with technical knowledge, external tools and best practices.
AWS also introduced OpenSearch Agent skills, a set of composable skills that bring OpenSearch knowledge into tools such as Kiro, Cursor and Claude Code.
Agent Skills enables AWS applications to connect natively to OpenSearch, meaning agents running on those applications are embedded in OpenSearch and can be monitored more effectively. This follows a trend among hyperscalers to help enterprises gain better visibility into how their agents are performing, said Lian Jye Su, an analyst at Omdia, a division of Informa TechTarget.
“All the hyperscalers acknowledge that the way machines access the cloud resources and how they are being monitored are very different from how human users would have been doing in the past,” Su said. “So, they’re making applications and their infrastructure a lot more open and a lot more native to agentic AI.”
The Challenge with Scaling
With its rebuild and new agent skills, AWS also tried to fix a major scaling problem that it experienced in previous versions of OpenSearch, which is that when users tried to use collections (a grouping of one or more data in a search workload) for agentic purposes, it was too slow and took about two minutes or more. With the rebuild, that has been reduced to seconds.
“Bringing that down to seconds is a big deal,” Patience said. He added that AWS is repositioning OpenSearch Serverless as a broader data layer for agentic systems. The cloud provider also committed to a roadmap to manage long-term memory agents and purpose-built search reasoning models later this year.
“It’s not just retrieval of information, but long-term memory and even search reasoning,” Patience said.
While AWS is really trying to provide a data infrastructure for agentic systems, organizations might still be hesitant about why they need it in their workflow.
“The agentic memory stuff is probably a little bit early, but I think it will appeal to teams inside of companies that are moving more rapidly along the agentic road,” Patience added.
