How should companies configure Guance Agent Teams?

It is recommended to configure settings based on job roles and scenarios, such as SRE Agent responsible for stability and fault handling, Security Agent for risk detection and response, FinOps Agent for cloud resource cost analysis, and Test Agent for quality assurance and regression verification. Each Agent can be configured with independent permission boundaries, knowledge bases, tool scope, and working context.

How is Guance Agent Teams memory managed?

Guance Agent Teams Supports conversational context and long-term knowledge accumulation. Session context is used to continuously analyze the same alarm, fault, or change; Long-term knowledge is used to accumulate historical incidents, response plans, runbooks, and best practices, and to control access to sensitive information through permission policies, data anxification, and audit records.

Guance Agent Teams How is the billing performed?

Guance Agent Teams Subscribe monthly by Agent at ¥999 per month / Agent. Each Agent subscription grants about 40,000 Credits per month, supporting about 500 complex tasks; Any excess is billed based on actual Credits consumed. Copilot Advanced Q&A shares the same Credits pool with Agent Teams task execution.

Will Guance Agent Teams replace SRE or DevOps teams?

Guance Agent Teams Better suited for handling repetitive, intensive, and time-consuming investigation tasks, such as initial alarm screening, log retrieval, indicator association, trace analysis, similar incident search, and review report generation. Business judgment, risk trade-offs, major change approvals, and cross-team coordination remain handled by engineers.

Guance Agent Teams How to control operational risks in production?

Agent Runs by default with read-only and least privilege. Administrators can configure tool whitelists, skill approvals, action grading, access scope, data anonymization policies, and high-risk operation approval flows, while maintaining complete operation records and evidence chains.

Can enterprises customize their own observable Agent?

Yes, you can. Enterprises can configure their own Guance Agent based on team processes, toolchains, permission policies, and knowledge bases, such as on-call analytics Agent, release stewardship Agent, cost optimizationAgent, or management summary Agent.

Guance Agent Teams Can you work 24×7 hours a day?

Yes, you can. Agent Suitable for long-term tasks such as continuous inspections, initial alarm screening, change observation, anomaly attribution, and periodic reporting. Companies can set trigger conditions, execution frequency, and budget boundaries based on scenarios.

Guance Agent Teams Can you connect to existing DevOps toolchains?

Yes, you can. Through MCP Server, OWL CLI, OpenAPI, and A2A collaboration, Agent can access Kubernetes, Terraform, CI/CD, tickets, Slack, Feishu, DingTalk, and existing enterprise automation platforms.

If AI Agent misjudges, how can it be traced and corrected?

Guance Agent Teams Retains analysis basis, tool calls, parameters, results, approval records, and execution feedback. Teams can pinpoint the source of problems based on the chain of evidence and deposit corrected runbooks or knowledge rules back into the system.

Guance Agent Teams How to go online?

It is recommended to start with clear scenarios such as duty alarm analysis, core link fault location, posting change observations, optimizing cloud resource costs, or generating daily and weekly reports. After completing role selection, tool integration, permission configuration, and approval strategies, gradually expand to more services and teams.

Why Agent Team instead of a regular AI assistant?

Ordinary AI assistants mainly answer questions; Agent Teams emphasize role division, tool invocation, process collaboration, and result delivery. Different Agent can separately handle tasks such as alarm analysis, trace association, release change comparison, security checks, and cost optimization.

Are Guance Agent Teams's operations auditable?

Auditable. Agent tool calls, skill usage, parameter input, data sampling, inference processes, approval results, and execution actions can all form a chain of evidence, facilitating security audits, fault reviews, and compliance checks.

What is the difference between Guance Agent Teams and the Universal Agent framework?

The Universal Agent framework is better suited for building personal assistants, self-hosted toolchains, or customized automation projects. Guance Agent Teams Targeting enterprise-observable scenarios, it features built-in metrics, logs, links, RUM, topology, alerts, and change context, emphasizing permission whitelists, approval flows, audit traceability, data governance, and team-level governance capabilities.

Which teams are not suitable for direct use of Guance Agent Teams?

If a team only wants to do AI demonstrations without clear alert analysis, fault location, release protection, cost optimization, or security response scenarios, short-term value will be limited. Guance Agent Teams Better suited for teams that already have basic observable data, service catalogs, alert policies, and operations processes.

Agent Teams

Your dedicated AI team

24/7 to solve your problems

From data to decision-making, from analysis to execution, Agent Teams let AI Agent truly connect your production environment and become a powerful colleague for your team

Start free Learn more

Pending distribution frontend-proxy 5xx with SLO exceptions Lasts 32 meters

Entry services surged by 5xx, SLO and Trace triggered alarms at the same window, awaiting takeover.

On dutySRE duty group Latest operationTriggering new faults handlerNot allocated

Processing checkout-api CrashLoopBackOff Lasts 18 meters

Pod restarted 12 times, the most recent exit code is 1, and the log points to an abnormal Redis connection configuration.

On dutyApplication on duty handlerYuki

Pending distribution Payment-service latency increases Lasts 42 meters

P95 latency is 2.8 times higher than baseline and has not yet triggered a serious SLO alert.

On dutySRE duty group handlerNot allocated

Important 06/30 12:55:00 (32 minutes ago) frontend-proxy 5xx with SLO exceptions

Lasts 32 meters

Pending distribution Handler: Not assigned

Fault details Collaborative records Operation records

Current context Wait for AI to inherit the alert context

The current fault has been carried into service, resource, namespace, timerange, and event ID.

Event Four have not been restored

Trace P99 9534ms

SLO 0.5258%

Read the fault context query_incident

State Circulation Pending → → in processing has been resolved

Abnormal signal: The fault center has generated a fault pending assignment

Build your own AI team and solve complex problems with ease

Create a dedicated team Agent

Quickly create dedicated Agent based on team roles and business scenarios. Configure knowledge, tools, and permissions to help Agent understand your business better.

Operations and maintenance experts

Performance analyst

SRE assistant

Custom Agent

Deploy it into your production environment
Agent that can truly carry out tasks

Securely connect to your production environment, provide system monitoring and tools, perform real tasks such as troubleshooting, analysis, and changes, and solve problems in a closed loop.

Production environment connection diagram

Connected to Guance
It can help you further explore the issue

Deeply integrated Guance, accessing metrics, logs, links, and other data, guiding you to further analyze root causes and provide actionable recommendations.

IndicatorsCPU · Memory · Load

Host informationorder-service · healthy

EventNearly 1 hour · No abnormalities

Developer

How has the @order-service service been operating in the past hour? Are there any abnormalities?

Coding Agent

Metrics, events, and host information are being correlated...

Current status confirmed

CPU5.39%

Memory46.17%

Load0.46

Disk73.6%

Judgment:
All indicators are within healthy ranges,
No abnormal alarms found.

Build Channels to connect Agent to familiar work platforms and become your colleagues

Seamlessly integrate platforms like Slack, Feishu, DingTalk, and WeCom to collaborate with Agent in your familiar environment.

Slack

Sending a flying letter

DingTalk

WeChat Work

More platforms

Collaborate with Coding Agent
Helps R&D engineers quickly analyze, locate, and fix online issues

Collaborate deeply with Coding Agent to automatically analyze issues, locate code, generate remediation solutions, and accelerate the closed loop of online issues.

const service = OrderQueryService.getDetails(orderId);trace.query("orders").where({ status: "paid" }); SELECT * FROM orders WHERE user_id = ? LIMIT 50;span.setAttribute("db.duration", 1240); if (cache.miss) await repository.findOrderItems();metric.histogram("order.p95.latency").record(1800); await Promise.all([fetchPayment(), fetchShipment()]);logger.warn("slow sql detected", queryHash); OrderQueryService.getDetails -> queryOrderSkuListdb.statement: SELECT sku_id, price, count FROM order_item apm.linkTrace(traceId, serviceName, endpoint);slowSql.count = slowSql.count + 1; pipeline.scan(productionTraffic, errorLogs);duration.breakdown.database = 0.67; rootCause.rank(["database", "cache", "network"]);agent.context.attach(observabilitySnapshot); const p95 = quantile(latencySamples, 0.95);return composeOrderDetails(order, items, user); ordersRepository.findById(orderId, { traceId });db.pool.waiting = connectionPool.pendingCount; trace.child("mysql.query").setStatus("slow");cache.hitRate("order.details").below(0.42); SELECT order_id,total,status FROM order_summary;agent.read(span.events).filter(e => e.error); service.map("checkout").edge("mysql-primary");slowSql.topK(3).groupBy("statement_hash"); if (latency.p95 > 1500) raiseFinding("P95");profiling.cpu.attach("OrderQueryService"); queryPlan.rows_examined = 184205;indexAdvisor.suggest(["idx_order_user_time"]); timeline.mark("db", 0, 1206).mark("app", 1206, 1800);log.pattern("timeout").count(328); span.link(logs, metrics, traces, profiles);rootCause = "database query fan-out"; agent.summary.write(performanceFindings);patch.plan("add composite index").estimateImpact(); observe.prod("order-service").window("15m");notify.channel("incident-room").preview();

Production Traffic1.24M Requests

P95 Latency1.8s

Database Time67%

Error Logs328 Events

Developer

Analyze the actual operation of OrderQueryService online

Coding Agent

Currently correlating production environment observation data...

Production environment analysis completed

Order service P95 latency reaches 1.8 seconds,
Database queries account for 67% of total time consumed,
Discovered 3 high-frequency, slow SQL entries.

The main performance bottleneck lies in
OrderQueryService.getDetails()

Supports automation
Let Agent work 24/7 without rest

Through automated rules and task scheduling, Agent continuously monitor, automatically process, and generate reports to ensure stable system operation.

Trigger conditions (Alert/Timing/Event)

Automatic analysis

Carry out tasks

Notification feedback

Completing the closed loop

Build your Agent team now

Let AI Agent truly become a super colleague on your team

Start free Book a demo

Agent Teams

The status changes to

Build your own AI team and solve complex problems with ease

Create a dedicated team Agent

Deploy it into your production environmentAgent that can truly carry out tasks

Connected to GuanceIt can help you further explore the issue