TLDR: At OpenSearchCon Europe 2026, Carl Meadows announced three major shifts positioning OpenSearch as the foundation for agentic AI: OpenSearch Launchpad for natural language application deployment, automated hybrid search configurations, and high-cardinality telemetry tracking. Additionally, CERN joined the OpenSearch Software Foundation to handle massive scale workloads. |
Based on the OpenSearchCon Europe 2026 keynote by Carl Meadows, Director of Product, Amazon OpenSearch Service | Watch the full keynote →
The phrase “agentic AI” has become one of those terms that means everything and nothing, depending on who is using it. At OpenSearchCon Europe 2026, Carl Meadows, Chair of the OpenSearch Software Foundation Governing Board and OpenSearch Product Lead at AWS, gave it some actual shape.
His keynote was less about the concept of agentic AI and more about what it concretely changes for search infrastructure: who gets to build production search applications, how fast they can do it, and what the observability layer for AI-powered systems needs to look like. Three announcements anchored the talk, and together they represent a meaningful shift in what OpenSearch is becoming.
OpenSearch Launchpad: From idea to running application in minutes
The most immediately striking demo of the keynote was OpenSearch Launchpad. Meadows showed a fully functional hybrid search application being built directly from an AI integrated development environment using natural language—no deep search expertise required. The application came up in minutes.
That is a significant change in who can build with OpenSearch. Historically, building a production-grade hybrid search system required knowing the query DSL, understanding how BM25 and vector scoring interact, and working through a meaningful amount of configuration. Launchpad abstracts that initial complexity so that developers who are not search specialists can get to a working application and then tune from there. The details of how far that abstraction goes and where it hands back control to the engineer are worth watching the full demo to understand.
The OpenSearch Relevance Agent: Closing the loop on search quality
The second announcement was the OpenSearch Relevance Agent, a tool that helps teams improve search quality over time using real user behavior data—all from within OpenSearch Dashboards. This is meaningful because search relevance improvement is typically a painful, manual, expertise-heavy process. You need to know what users are clicking, what they are ignoring, and how to translate that signal into changes to your scoring logic.
The Relevance Agent is designed to close that loop more continuously and with less manual intervention. It builds on the User Behavior Insights (UBI) work the OpenSearch community has been developing and brings it into a more integrated workflow. For teams running production search where relevance quality directly affects business outcomes, this is the kind of tooling that moves the needle.
The Observability Stack: Open source catches up
The third piece Meadows highlighted was the OpenSearch Observability Stack, a single-click deployment that brings logs, distributed tracing, and Prometheus metrics together in one unified experience. What made this notable was the framing: this level of integrated observability has historically been available only through commercial vendors. OpenSearch is making it accessible as open source.
For organizations running OpenSearch at scale who also want a coherent observability picture without paying for a proprietary stack, this is a meaningful development to evaluate.
A growing foundation
Meadows also used the keynote to welcome new members to the OpenSearch Software Foundation: CERN, Big Data Boutique, OpenSource Connections, and Resolve Technology. CERN joining as an associate member, an organization that processes petabytes of physics data from the Large Hadron Collider, is a signal worth paying attention to. It reflects the direction the project is moving: toward handling the most demanding data workloads in the world as open source.
The through-line of the keynote was that the OpenSearch community’s open-source foundation, combined with its vector database capabilities, positions it well for a world where AI agents need a reliable memory and retrieval layer. Whether agentic AI lives up to its promise is still playing out, but the retrieval infrastructure it depends on is being built now, in the open.
