IBM Instana ships AI agent observability to preview
IBM has released a public preview of AI agent and LLM observability capabilities in Instana, its application performance monitoring platform, extending visibility into agent-driven workflows and decision-making at scale.
The announcement was made at IBM Think, where the company positioned the new observability module as a tool for organizations deploying autonomous agents and generative AI systems. IBM's materials describe the addition as part of recent Instana innovations aimed at helping teams understand application performance across AI-powered infrastructure.
What is being previewed
Instana's new observability layer covers prompt tracing, token consumption monitoring, agent decision tracking, and dependency mapping across agentic workflows, according to third-party analysis. IBM states: "Recent Instana innovations include dedicated observability for generative AI applications, large language models (LLMs) and AI agents, helping organizations understand application performance, prompt flows, token consumption, dependencies and operational behavior across AI-powered systems."
The capability is positioned alongside IBM's broader agentic platform, which includes IBM Concert, the company's enterprise-grade agent orchestration framework. Together, the observability layer and Concert aim to address the operational complexity of running agents in production—a growing concern as enterprises move beyond single-agent chatbots to multi-agent systems handling critical workflows.
Why it matters now
Agent deployments are accelerating across banking, healthcare, and customer service sectors, but visibility into agent behavior remains fragmented. Most monitoring platforms were built for traditional microservices and containerized workloads, not for systems where agents make autonomous decisions, call APIs, and consume LLM tokens at unpredictable rates.
Instana's observability for agents addresses three specific pain points: cost control (tracking token usage per agent and workflow), reliability (tracing decision pathways and failures), and compliance (logging prompt inputs and agent actions for audit trails).
Product status
The feature is available as a public preview, meaning organizations can test it in non-production environments and provide feedback before general availability. IBM has recognized observability as a critical market, having been named a leader in the 2026 Gartner Magic Quadrant for observability platforms, which evaluates vendors on execution and vision.