All Products
Search
Document Center

Application Real-Time Monitoring Service:[Platform upgrade] Upgrade of LLM Application Monitoring to AI Agent Observability

Last Updated:Jul 01, 2026

AI Agent Observability is a module of Application Real-Time Monitoring Service (ARMS) that provides observability for AI Agents, AI applications, models, and tools. This module was previously known as LLM Application Monitoring.

Upgrade timeline

Time

Changes

June 1, 2026

Data migration is completed silently.

July 30, 2026

LLM Application Monitoring is renamed to AI Agent Observability. The console access path is routed to Cloud Monitor 2.0.

Key capabilities

AI Agent Observability builds on its existing features by adding the following core capabilities:

  • Global topology and health monitoring — View the global topology, dependencies, and health status of AI applications, AI agents, models, tools, and other observable entities in real time.

  • Trace analysis — Native trace views for agent reasoning and execution paths, including trace trees and trace diagrams. Multi-modal data is natively supported for preview.

  • Session analysis — User-perspective session views that reconstruct user-to-agent interactions, with support for multi-turn conversations and long-running sessions.

  • Scenario-specific dashboards — Dedicated views that cover token usage, model performance, tool invocations, and user behavior analysis.

  • Alerting — Alert metrics for AI agents and AI applications across model invocations, tool invocations, token consumption, and agent self-invocations. Intelligent analysis and root cause identification are supported for agent anomaly alerts.

Impacts

  • Data migration: User data from LLM Application Monitoring is automatically and seamlessly migrated. No code or configuration changes are required.

  • Console entry: The LLM Application Monitoring menu is renamed to AI Agent Observability. The console access path is routed to Cloud Monitor 2.0.

  • Billing model: This upgrade does not affect the billing model.