Autonomous operations, starting with data.
Dashboards tell you something is wrong. Varqeni tells you what, why and how it knows, does what you have approved, checks that it worked, and brings in a person only when judgment is needed.
watching
Above your telemetry, across your stack.
Every Varqeni agent works the same way, whatever it watches: a signal, an investigation with evidence, a decision checked against your policy, an action only where you allow one, verification, and escalation when the fix fails or the evidence runs out. Data, Infrastructure and Service Operations run in the lab today.
Data Operations Live in lab
Databases, change data capture, replication, Kafka Connect and the Kubernetes workloads behind them.
Reads today
PostgreSQL
Debezium
Debezium Server
Kafka Connect
Apicurio Registry
Kubernetes
Prometheus
Planned
PostgreSQL on any cloudManaged change data capture services
Infrastructure Operations Live in lab
Compute, containers, storage, networking, health checks, certificates and autoscaling, on any cloud or your own hardware. In the lab it watches the estate and runs four fixes after a named approval: a host's metrics exporter, a disk full of old logs, a hung workload and a bad rollout (32 of 32 closed-loop trials).
Reads today
Linux hosts
Kubernetes
PrometheusProxmox virtual machinesMinIO storage
Kafka brokers and topic freshness
AutoscalersLoad balancersTLS certificates
Planned
OpenTelemetry metricsPublic cloud compute
Service Operations Live in lab
Applications, APIs, dependencies, error rates, scheduled jobs and deployments. In the lab it watches service endpoints, the share of real requests that fail, which services call which, scheduled jobs and deploys. It names the cause, and blames a failure once, at the service where it started (42 of 42 blind trials), then rolls back a bad deploy, including one that fails only some requests, or restarts a hung workload after a named approval (32 of 32 closed-loop trials).
Reads today
HTTP endpoints, probed from outsideRequest error rates, from each service's own counters
Traces, through the OpenTelemetry CollectorDependencies between services, configured or learned from traces
Error logs and a service's own metrics, through the OpenTelemetry Collector
Scheduled jobs (CronJobs)
Deploys as suspects
Planned
Logs from other collectors
Custom operations Live in lab
Trigger→Investigate→Evidence→Policy→Action→Verify→Escalate
Your own runbooks as governed operations, inside the same guardrails as every Varqeni agent. Custom operations · All integrations
Four agents. One set of rules.
Built for the pipelines that cannot stop
Live pipeline health
Heartbeat age, replication slot position, connector and task state, and registry health, in one view.
Verified investigations
Real signals open an investigation. Evidence is gathered, and each finding is checked before it reaches you.
More than one pipeline
Watch several pipelines from one console, each through its own read-only access.
Tested against faults
Blind fault trials inject known failures to measure how well investigations find the real cause.
Governed actions
A short list of reviewed fixes, run after a named person approves the incident, or on their own inside a policy you write. Every one is verified and sealed.
Full audit trail
Signals, evidence, findings, approvals and actions, recorded with their times.
Monitoring built in
Dashboards that build themselves for every system it watches, with incidents and fixes marked on each chart, and Explore for any metric.
Every machine, one page
Resources lists the estate the way Proxmox does, with what each machine is given, what it uses and what runs on it.
Connections you can test
Every data source and integration in Settings, checked before it is saved, versioned, and rolled back in one step.
Works with
Put Varqeni on your pipeline
Varqeni is in early access. Tell us about your pipelines and we will set up a trial around them.