OpenSearch Introduces New Version of Observability Stack

OpenSearch has launched version 3.6, which is the project’s first long-term support release. This update adds a new OpenSearch observability stack, application performance monitoring, and a set of searches, along with the agent tools designed for developers and operations teams. The OpenSearch launchpad is a major addition because it is a guided tool that assists users in building search applications from sample documents and inputs entered by users. By using conversational workflow, it can offer a local setup, choose retrieval approaches, configure models, and develop a working user interface.

Version 3.6 also adds the OpenSearch relevance agent as an experimental feature. It is developed to automate search relevance tuning, analyze user behavior, create hypotheses, and evaluate changes with the help of the natural-language interface in OpenSearch dashboards. The release provides an experimental multi-agent orchestration platform for relevant agents and various specialized agents. It also has a unified agent registration API that combines various manual setup guidance as a single API call. It further provides a new conversational_v2 agent type with standardized inputs designed for documents, text, and message history, along with several images.

According to Precedence Research, the observability tools and platforms market size accounted for USD 11.80 billion in 2025 and is predicted to increase from USD 13.40 billion in 2026 to approximately USD 49.60 billion by 2035, expanding at a CAGR of 15.44% from 2026 to 2035. The market is significantly growing due to the increasing integration of AI/ML in observability tools to automate operations in leading sectors, the need to manage distributed trading, and integration by the BFSI and healthcare sectors.

Observability is a major part of this new release. The OpenSearch observability stack combines OpenTelemetry collector, Data Prepper, OpenSearch, Prometheus, and OpenSearch dashboards in a preconfigured deployment. All these can be launched with the help of Docker Compose or an installer. It is intended to monitor microservices, web applications, and AI agents. It includes service maps, distributed tracing, log analytics, Prometheus-compatible metrics, and other tools for tracking large language model calls, execution graphs, and token usage.

Users can further group services with attributes like SDK language or team, filter by errors or fault rate thresholds, and inspect time-series charts for requests, latency, and errors. Agent traces were also added, aiming to monitor generative AI applications via opentelemetry-based instrumentation.