Log intelligence from inside your Kubernetes pods.
Axiona is a self-hosted platform that collects, stores and searches the logs of your enterprise infrastructure, including the logs inside your pods. It raises the alerts, helps your team investigate, and needs no agents.
Collection that reaches inside the pod
Axiona
Collect, normalize, store and search in one self-hosted platform.
What your team does with it
Search logs, act on alerts, run incidents, and rely on stored evidence that cannot be silently replaced.
Inside the pod
Axiona collects logs from inside the pod itself, where your application writes them.
Agentless
No agent on your Linux hosts, and no sidecar or DaemonSet in your clusters.
Self-hosted
Axiona runs in your environment, so your logs stay with you.
From collection to investigation, in one platform.
Inside the pod
Reads application files inside a running container through the Kubernetes API, as well as stdout and stderr. No sidecar and no DaemonSet. Collection follows a label selector across Pod restarts and rescheduling.
Agentless on Linux
Log files are collected over SSH and SFTP with a pinned host key. No agent, no daemon and no package on the host.
Never loses its place
Durable checkpoints survive the loss of a collector worker, and rotation and truncation are handled without duplicate or lost events.
An independent alert engine
Nine rule types, including windowed counts, event rates and absence ("this heartbeat stopped"). One recurring condition is one alert with an occurrence count, not a thousand notifications. Test a rule against a sample event before you save it.
Alerts with a lifecycle
New, active, acknowledged, resolved and closed, plus suppressed. Illegal transitions are refused, and a recurrence never overwrites a human's acknowledgement.
Maintenance-aware suppression
Recurring maintenance windows run in their own timezone. Planned work stops paging people, but every event stays searchable and audit-complete.
Incidents people own
People open incidents, and correlation links matching alerts into them, marked permanently as automatic links. Ownership, notes and a full audit trail.
Search and an operator console
Full historical search through a controlled API, filtered on the server by each person's system grants. Every event is normalized into one typed contract first.
Optional local AI
Alert analysis and incident summaries from a model that runs inside your deployment, on a network with no route out. Credentials and PII patterns are stripped from prompts. With AI off, everything else works.
Access control that knows your systems
Permissions answer what a person may do and which systems they may do it to. Custom roles, privilege ceilings, and an audit trail that cannot be rewound.
Backup and recovery built in
Scheduled, encrypted and verified backups. A restore needs a second administrator, takes a safety snapshot first, and cannot erase the audit trail.
Offline licensing
Signed licences are activated offline. Axiona never contacts a vendor service in normal operation, and it degrades gracefully instead of stopping.
Secure by design, and audited before release.
A full audit, with six static scans, dependency audits on every running container and live checks, was run and closed with zero open Critical, High or Medium findings. These are the highlights.
TLS at the edge
HTTPS only by default; plaintext is an explicit opt-in. The full security header set and a strict content security policy are on.
Every service authenticated on the message bus
Each service is its own NATS user with a subject allow-list. A compromised collector cannot publish an alert state change.
A least-privilege log store
Six scoped OpenSearch accounts. The writer can create documents and nothing else: no read, no update, no delete. Stored evidence cannot be silently replaced.
Network segmentation
Four Docker networks, and only the edge network has a route out. The AI runtime has no internet access at all.
Credential isolation
A source credential can resolve only to a per-host secret or to an explicit shared allow-list. It can never name a platform secret.
Bad events cannot block good ones
An invalid event is dead-lettered with its payload and acknowledged on its own, so it cannot fail the batch around it.
Dependencies audited
pip-audit is clean across all ten service containers, and the production npm audit is clean.
Check Axiona against your environment.
Supported versions and platforms depend on the release. Email us what you run and we will send the compatibility and supportability matrix as a one-page flyer.
See Axiona on your clusters.
Email us the clusters and log sources you work with. We reply to arrange a walkthrough.