Asset
Hosts, databases, cloud resources and clusters are recorded with their project and relationships. Jobs pick their targets from these records.
Job
Reviewed scripts and templates run across the targets you pick, in a batch or on a schedule.
Approval
Confirmation or approval follows risk and team policy. An approval is bound to this run’s command, parameters and targets, admits it once and expires after 30 minutes by default.
Audit
Key actions leave an audit record of who acted, when, on which resource and with what result, exportable by time range.
Private production operations control for the enterprise
Production operations
on a controlled path
ChronoOps runs inside your own environment. Jobs run against the targets in your asset records, confirmation or approval follows risk and team policy, and key actions leave an audit record. Cloud bills are attributed to projects, idle resources are flagged and optimization suggestions follow. AI helps diagnose and drafts a response; a person reviews it before anything runs.
Community: free licence, up to 50 nodes · Both editions run in your environmentProduct UI · Synthetic data
Asset overview and to-do counts
The overview opens with managed, active and stale asset counts: 142, 135 and 7 in this synthetic data. Host sources are counted separately by Agent, cloud sync and manual entry.
- 1Total managed assets, covering hosts, application services and cloud resources
- 2Quick actions with to-do counts for cloud sync failures, asset relations awaiting review and host identity conflicts
- Cloud connections
- Relationships
- Database inventory
Cluster inspection results and history
The cluster workspace inspection tab, with the latest verdict, the healthy rate, a per-run trend and the inspection history on one page.
- 1Latest inspection: needs attention, 2 issues found, 63% healthy over the last 8 runs, full coverage; each cell in the trend strip below is one run
- 2Four cluster workspace tabs: inspection, resources and diagnosis, operations and maintenance, logs and events
- Resources and logs
- Inspections
- Operations
Inspection report graded against thresholds
The report's verdict is critical, with a reminder that a successful run does not mean the checks are healthy.
- 1Critical and unclaimed: a successful run does not mean the checks are healthy; claim it or link a re-check
- 2Checks on two hosts, each compared with its thresholds: critical, needs attention, pass
- Scripts and templates
- Scheduling
- Reports
Remote access sessions and recordings
The access workspace lists six sessions, each with its target, session status, risk level and recording status; some can also be connected to and replayed.
- 1Each session's target, session status, risk level and recording status
- 2Details, connect, replay and audit links for sessions
- SSH
- SFTP
- RDP
- VNC
See build, deployment and latest release by service
The table lists, per service, the repository and build tool, the deployment environment, the target hosts and the latest release status; the first three rows read waiting for confirmation, deploying and succeeded.
- 1Release workspace task list with Prepare release selected
- 2Repository and build tool, environment, target hosts and latest release status for the first two services
- Service onboarding
- Builds and releases
- History
Alert detail and per-notification delivery results
A critical alert pushed by Alertmanager, still firing.
- 1Alert pushed by Alertmanager: status, labels and the 确认接手 (take ownership) button
- 2Result of each notification: Feishu and DingTalk delivered, email failed
- Alertmanager
- Rosters and escalation
- Timeline
AI-assisted diagnosis of an alert event
The diagnosis record of one alert event. The AI lists possible causes and marks every suggested action as needing human review; the runbook-knowledge card answers with three knowledge-base citations.
- 1AI root-cause analysis: possible causes and suggested actions that need human review
- 2Runbook knowledge: an answer with citations from knowledge-base documents
- Alert events
- Process samples
- Read-only
Estimated and confirmed savings kept apart
The Cloud costs page's savings panel shows realized savings confirmed by Finance on their own at the top; estimated monthly savings come below, split by recommendation status.
- 1Realized savings confirmed by Finance, shown on their own: 3 recommendations, CNY 2,786.40
- 2Estimated monthly savings, listed by recommendation status
- Billing
- Budgets
- Recommendations
Connect the systems you already use
- Kubernetes
- Linux / Windows
- Prometheus
- Loki
- OpenTelemetry
- Enterprise SSO
Two loops
for execution and cost
One governs production operations, from asset records to the audit trail. The other governs cloud spend, from the bill to optimization suggestions. Every step leaves a record in the platform that you can check later.
Cloud cost optimizationEnterprise feature
Which project the spend belongs to, what sits idle, what could be saved
Bill
Public-cloud bills sync automatically by billing period into one place.
Attribution
Rules attribute costs to projects and resources; anything left unattributed is listed on its own.
Idle detection
Unattached disks and unbound elastic IPs are flagged for on-call review. This covers Volcengine today; other clouds are confirmed during evaluation.
Optimization suggestions
Suggestions come with estimated savings, and the platform never releases a resource on its own. What was actually saved is confirmed by later bills and finance.
Your everyday operations
Without switching tools
Key actions leave an audit record, exportable by time range, up to 10,000 rows at a time.
Assets
Assets & CMDB
Connect cloud accounts on AWS, Azure, Google Cloud, Alibaba Cloud, Tencent Cloud, Huawei Cloud, Volcengine and OpenStack; those resources sit in one inventory with your hosts, databases and applications. Each record carries the project it belongs to and what it relates to, and links to the matching diagnosis or operation page.
Read the guideKubernetes
Resource status, Pod logs and events for clusters, namespaces and workloads sit in one workspace. Inspections find problems on their own, and the day-to-day operations the product supports run here too.
Read the guideOperations
Jobs
Reuse reviewed scripts and templates for batch jobs and scheduled inspections. An approval covers one run and expires after 30 minutes by default. A finding gets an owner, a record of the follow-up and a re-check.
Read the guideBastion
Connect through ChronoTerminal, managed SSH, file transfer and remote desktops. Accounts, access and session records are managed together, and recordings expire after 30 days by default.
Read the guideReleases
Builds, releases and change history are organized by application and service. Follow a task, check what it did, and roll back where a rollback operation exists.
Read the guideObservability and AI
Observability & on-call
Alerts come from monitoring you already run, Prometheus Alertmanager and others, routed by label to Feishu, DingTalk or email. The event page shows whether each notification got through. Rosters, overrides, escalation, silences and maintenance windows are all there. An on-call engineer acknowledges in the platform and can work against the unified timeline of recent jobs, releases and bastion sessions.
ChronoOps does not rebuild your monitoring or replace phone and SMS paging tools.
Read the guideAI diagnosis
For an alert event it suggests likely causes and next steps, quoting your knowledge base where there is a match. Every conclusion from a process sample cites the facts collected and says what cannot be determined. Suggestions are read-only: nothing changes production on its own.
Included in Community and Enterprise. Connect a model service first, such as an enterprise model gateway or private model. Alert-event diagnosis works on alerts received through Observability & on-call.
Read the guideCloud costs
Public-cloud bills come together in one place, with costs, budgets and recommendations broken out by project and resource. Estimated savings stay apart from what a later billing period confirms.
Read the guidePrivate & localized deployment
Runs in your environment, with private-network and offline delivery. Localized-stack adaptation covers domestic operating systems, chip architectures and databases; the combinations are confirmed for the target release and environment.
AI helps diagnose
People review before it runs
Alert events arrive through Observability & on-call. AI suggests likely causes and next steps, quoting your knowledge base where it matches. For process samples it cites only the facts collected and marks what it cannot determine. Execution still needs permission checks and approval under your team’s rules; an AI-generated remediation plan needs at least two different approvers.
- Diagnosis with evidence
Process diagnoses show their evidence, and say what is missing
- A plan with its rollback
A generated remediation plan lists its steps, risk level and rollback
- Controlled execution once authorized
Only predefined actions and job templates run. You explicitly authorize a low-risk single action, everything else goes to approval, and the result is read back afterwards
AI never changes production on its own: nothing it proposes runs without a person’s authorization or approval.
In both editions. It works once you connect a model service — an enterprise gateway or a private model — and key actions are audited.
Visit the Trust Center
From diagnosis to actionIllustrative workflowFinding: disk capacity needs attention
Compare host state with check results and state the impact.
Recommendation: choose a suitable response
Write out the evidence, the target and the limits, and keep guesses out of the conclusion.
Authorization: confirm before execution
Confirm or approve according to risk and team policy. An approval is single-use: 30 minutes by default, 15 for high-risk definitions.
Verification: inspect the result
The execution record stays. Run another check or follow up if you need to.
Start with the problem
in front of you
You don’t need to connect everything at once. Start with cluster inspections, remote access or batch jobs, and see whether they fit.
Describe your scenario and book a demoA cluster needs attention
Start from an inspection finding, open the resource, read its Pod logs and events. Then pick the action that fits.
- Inspection
- Diagnosis
- Operation
Related modulesKubernetes
Routine work keeps repeating
Turn the scripts you keep rerunning into templates and run them across the targets you pick. An inspection finding gets an owner and stays tracked through its follow-up check.
- Template
- Scheduled job
- Follow-up
Related modulesJobs
Cloud bills don’t match your resource records
Put your asset records next to the cloud bill, read costs by project and resource, then read the recommendations. Whether anything was saved, later bills tell you.
- Asset
- Cost allocation
- Evaluation
Related modulesAssets & CMDBCloud costs
Your environment
Start with a small trial
Start with a small set of resources before expanding. We will confirm which systems to connect, how to deploy and what implementation involves.
Community
Core operations features, under a free license.
- 50
- managed nodes
- 1
- Kubernetes cluster
- One tenant, core assets, operations and AI diagnosis
- Job templates, approvals and in-platform audit
- Software packages provided; source code is not open
Enterprise
For teams managing more resources and access permissions.
- 50+
- managed nodes, added by license
- Scoped
- features confirmed during evaluation
- Add managed nodes and enterprise features by license
- AI diagnosis, cloud costs, localized-stack adaptation assessed per environment
- Implementation, training and agreed support
Private deployment is supported. Network, offline and platform compatibility are confirmed during evaluation.
Deployment detailsFrequently asked questions
Can’t find your question? Email us directly
Do we need to replace our monitoring, logging or ITSM?
No. ChronoOps plugs into the monitoring, logging and other operations systems you already run, so troubleshooting shows the related assets, alerts and changes. What to connect and how depends on your data sources, permissions and network; we settle that in the demo.
Does every action need approval? Can one person use it?
No. Routine remote access proceeds once multi-factor authentication passes; a trusted device skips the MFA prompt at login for 7 days, but sensitive actions such as remote access verify again. What needs confirmation or approval follows its risk and your team’s policy; a job approval is single-use, 30 minutes by default. One person can run it alone, with the rules set at deployment.
Will AI make changes to production on its own?
No. AI diagnoses the problem and proposes a fix with its rollback; it never changes production on its own. Carrying a fix out goes through predefined actions or job templates only: permissions are checked, then a person explicitly authorizes it or it is approved under your team’s rules, an AI-generated remediation plan takes at least two distinct approvers, and the result is read back afterwards. Asset management, jobs and inspections work with no model service connected.
Can we deploy on a private network or offline?
Private deployment is supported, and delivery can be offline. Backend and frontend images are built for linux/amd64 and linux/arm64; the host agent is a binary, built for linux/amd64, linux/arm64 and windows/amd64. Operating systems, chip architectures, databases, external integrations and localized-stack environments are assessed item by item for your target release and environment.
Is Community Edition open source? How do we start?
Community Edition isn’t open source — you get free licensed packages. One tenant, up to 50 managed nodes, one Kubernetes cluster, and AI diagnosis as soon as a model service is connected, whether that’s an enterprise gateway or a private model. Request a trial with the form below. For cloud costs or other Enterprise features, just ask.
Book a demo
Explore the features you need
Tell us about your environment and the features you want to see. We will prepare a relevant demo and contact you within one business day.
- No production accounts or sensitive data required
- Request a Community trial or discuss Enterprise
- Discuss private deployment, offline and platform requirements