STAROps is a full-domain intelligent O&M platform that Alibaba Cloud built on large language models and AI agent technologies. It deeply integrates cross-domain observable data with large language model reasoning to overcome the limitations of traditional O&M tools, such as a high barrier to use and severe data silos. You define goals in natural language, and O&M agents autonomously complete a full closed loop of dynamic planning, safe execution, and result verification. STAROps fundamentally transforms O&M from passive response to proactive autonomy, giving enterprises 24/7 autonomous operations that protect the resilience of your business systems in real time.
What does STAR stand for?
Each letter in STAR represents a core design principle of the platform:
-
S — Sense: holistic observability. STAROps collects cross-domain data such as logs, metrics, traces, and topology through an observability model (UModel), breaking down data silos. It builds a real-time digital twin of your running system and provides complete context for AI-driven decisions.
-
T — Target: goal-oriented operations. Instead of reacting to alerts, STAROps focuses O&M on achieving goals. Define your O&M goals in natural language. The platform continuously evaluates deviation from those goals, pinpoints the factors that affect them, and drives decisions and execution.
-
A — Autonomy: autonomous operations. Multi-agent coordination drives a full sense-decide-execute-verify loop without step-by-step human intervention, enabling 24/7 autonomous operations. High-risk actions retain Human-in-the-Loop (HIL) approval, balancing efficiency with safety.
-
R — Resilience: business resilience. STAROps shifts from reactive firefighting to proactive protection. Continuous inspection catches risks early. When incidents occur, the platform automatically executes remediation — scaling, rollback, or traffic shifting — and verifies the results, shortening Mean Time to Recovery (MTTR). The platform captures operational experience in a knowledge graph, continuously improving system resilience.
Core capabilities
STAROps provides three core capabilities:
AI Chat
Use natural language to analyze alerts, query data, interpret metrics, and query and analyze logs. AI Chat turns tedious work on the command line into instant insights and significantly lowers the barrier to O&M work. Start a session with the intelligent assistant from multiple entry points, including STAROps, CloudMonitor, and Simple Log Service (SLS).
Missions
A Mission is a human-machine interaction mode designed for long-cycle asynchronous execution. Align on a goal once, and STAROps plans an asynchronous O&M plan that spans days or months based on mechanisms such as schedules and events, and then executes the plan autonomously. Missions convert repetitive manual intervention into reliable automated processes and significantly improve O&M efficiency. Missions support scheduled and manual triggering. A built-in Human-in-the-Loop (HIL) mechanism ensures the safety of high-risk operations.
Digital Employees
A digital employee (SRE Agent) is the intelligent execution entity of STAROps. Build enterprise-specific SRE agents with custom responsibilities, permissions, tools, and skills to build an O&M assistant that fits your business scenarios — reducing customization costs and accelerating R&D and operations productivity. A digital employee is both the counterpart that you converse with in AI Chat and the executor of Missions.
Core strengths
|
Benefit |
Description |
|
Unified data platform |
Alibaba Cloud's unified observability data foundation stores logs, topologies, metrics, and traces together. Supports petabyte-level daily ingestion, EB-level storage, and seconds-level analysis over hundreds of billions of records. Multi-zone deployment delivers 99.99% reliability. |
|
Operations digital twin |
Built on UModel, the digital twin uniformly models applications, services, resources, topologies, alarms, and change relationships. It supports custom extensions, real-time topology inference, and causal analysis. |
|
Data analytics operators |
General-purpose data analytics operators and observability AI analysis operators cover metric outlier detection, log clustering, trace analysis, performance profiling, and change backtracking. These operators improve root cause analysis (RCA) timeliness and reduce model inference costs. |
|
Flexible integration solution |
Provides multiple integration solutions such as OpenAPI, page integration, and IM integration (DingTalk and Feishu) to flexibly integrate existing workflows. |
Security and compliance assurance
-
Fine-grained authorization policy: Hierarchical RAM role authorization for operators and digital employees manages what people can do and what agents can access, enforcing least-privilege access and reducing the risk of unauthorized operations.
-
Manual intervention: Connect tools via MCP with Human-in-the-Loop (HIL) configured. High-risk write operations and dangerous commands require manual confirmation, and a blocking engine intercepts abnormal operations to prevent misoperations and malicious behavior.
-
Agent behavior audit: Retains complete records of conversation history, runtime artifacts, tool calls, CLI commands, and data access. Full-lifecycle agent behavior becomes traceable, repeatable audit evidence for compliance and security reviews.
-
End-to-end data encryption: All data in transit is encrypted via HTTPS/TLS. Observability data is encrypted at rest using KMS, and agent artifacts are likewise protected — ensuring data privacy and integrity throughout the pipeline.
Common scenarios
-
Scheduled intelligent inspection of Kubernetes clusters: automatically checks cluster health daily, generates structured reports, and compares historical differences.
-
Core service high-availability assurance: continuously monitors core services and automatically performs root cause analysis (RCA) when an alert is triggered.
-
Natural language-driven fault diagnosis: narrow down the troubleshooting scope through multi-turn dialogue and perform correlation analysis using UModel topology.
-
Regular data quality checks: monitors data pipeline health on schedule and automatically notifies you of exceptions.
-
Automated O&M report generation: aggregates operational data weekly or monthly and generates structured reports.
FAQ
Are session records, digital employees, and other resources created by a RAM user isolated from the Alibaba Cloud account?
No. Resource ownership for STAROps sits at the Alibaba Cloud account level, and STAROps does not isolate resources between a Resource Access Management (RAM) user and the Alibaba Cloud account. Session records, digital employees, and Missions created by a RAM user are still resources under the same Alibaba Cloud account.
To limit which resources a RAM user can view or operate on, isolate them by using a RAM access policy. For example, control the authorization scope of Actions such as cms:ListThreads, cms:GetThread, cms:ListDigitalEmployees, cms:GetDigitalEmployee, cms:ListMissions, and cms:GetMission.
Do you need to activate STAROps separately?
No. Click the STAROps entry point in the CloudMonitor 2.0 (CMS 2.0) or Simple Log Service (SLS) console. STAROps is then activated automatically by default, and no extra action is required.