OpenSearch 3.3’s New AI Agents Now Generally Available for Developers
Developed first as an Amazon-driven fork of Elasticsearch, OpenSearch, and more recently as a standalone community-driven alternative to Elasticsearch, OpenSearch has been moving fast and making new things.
As Natasha Woods, Senior Director of PR for the Linux Foundation told me in an e-mail exchange, “This release heavily focuses on being an all-in-one observability experience that brings logs, traces, and visualizations together in one place, removing the need for disparate technologies. New features in 3.3 enable smarter, automated visualizations, more powerful queries, sophisticated analysis, and faster insights from raw data.” These are all good things.
More specifically, OpenSearch 3.3, officially launched on Oct. 14, 2025, highlights OpenSearch’s aggressive eight-week update cycle. 3.3 represents the latest step in the platform’s evolution toward deep semantic search, advanced distributed tracing, and agent-driven AI workflows.
AI Agentic Search Reaches General Availability
The centerpiece of OpenSearch 3.3 is the graduation of AI agentic search and agentic memory APIs to general availability. This enables developers to deploy autonomous, plan-execute-reflect search agents directly within their data platforms. This means it can leverage AI agents to execute and refine search tasks by interpreting natural language questions and dynamically planning queries across data sources and tools. These agents can break down complex user questions, rewrite them for more intelligent retrieval, and synthesize directly relevant, summarized answers rather than just returning lists of plain document hits.
Semantic search is further enhanced with expanded relevance scoring and new controls for fine-tuning search results. That helps programmers to develop large-scale, custom AI inference tasks that are becoming increasingly common in generative AI applications.
In addition, the Machine Learning (ML) Commons plugin gains experimental batch inference support, enabling large-scale, distributed processing over vast vector datasets. Performance improvements from the new Seismic algorithm for neural sparse search also complement how OpenSearch deals with huge vector databases.
Semantic Search and Machine Learning Improvements
The major differences between OpenSearch 3.2 and 3.3 center around expanded capabilities for search, observability, and AI-driven workloads, alongside new features and optimizations in core components, especially for generative AI and memory-efficient operations.
The latest OpenSearch also focuses on experimental and user-facing enhancements. These include searchable snapshots across clusters, new ways to compare search results for analytics, support for multiple data sources in dashboards, and usability tuning for trace and security analytics workflows.
Unified Observability and Redesigned User Interface
On the frontend, OpenSearch 3.3 debuts a redesigned Discover interface in Dashboards. This features powerful new tools for log analytics and distributed tracing. Users can now compare search results interactively, integrate multiple data sources for richer insights, and monitor data transformation pipelines with processor chains.
Workload management has also been upgraded with rule-based auto-tagging, query monitoring, and expanded gRPC support. This also includes experimental streaming via Apache Arrow Flight. This makes OpenSearch more compatible with high-velocity data streams from modern observability stacks.
The release also introduces new system restrictions designed to protect cluster stability at scale. This includes limits on the maximum depth of nested JSON objects and property name lengths. This helps safeguard OpenSearch-based programs against hacks abusing data ingestion.
Upgraded Workload Management and Security Protections
In addition, security analytics see new connectors and granular control options, while distributed tracing and OpenTelemetry workflows benefit from improved dashboard instrumentation and trace analytics.
Sounds interesting? You can download OpenSearch 3.3.0 today for Linux (x64/ARM), Windows, Docker, FreeBSD, and Arch Linux. Whether you run OpenSearch on all cloud or on-prem, you’re covered.
