Stop Guessing. Start Observing.
When Something Goes Wrong in the Field, You're Flying Blind
Embedded devices don't get the luxury of kubectl describe pod. When something goes wrong, you're guessing from customer reports, shipping replacement units, or begging someone on-site to plug in a UART cable.
Koios gives you the observability stack your embedded devices have been missing. Real-time and historical telemetry from every device — even those running on a few kilobytes of RAM.
Two Pillars of Device Observability
Device Telemetry
Memory telemetry
Free heap, minimum free heap (high-water mark), heap fragmentation index, and stack usage per task.
Resource monitoring
CPU utilization per task, peripheral bus errors, watchdog timer resets, brownout detection events.
Reboot tracking
Classify reboot reasons (power loss, watchdog, panic, OTA, user-initiated) and track frequency over time.
Custom metrics
Define and report application-specific metrics — sensor readings, protocol error rates, connection retry counts.
Fleet aggregation
View metrics aggregated across your fleet, by firmware version, hardware revision, or region.
Threshold alerting
Set threshold-based or anomaly-based alerts on any metric. Get notified via webhook or email.
Log Streaming
Real-time streaming
Tail logs from individual devices in real time through the dashboard or API. No delay, no polling.
Full-text search
Search across all device logs. Find the needle in the haystack across thousands of devices.
Structured logging
Ship key-value structured log entries alongside free-text messages. Query on any field.
Deployment correlation
Cross-reference log spikes with firmware deployments, certificate rotations, or configuration changes.
Automatic context
Every log entry tagged with device ID, firmware version, device group, and timestamp automatically.
Retention policies
Configure how long logs are retained per device group or globally. Comply with data retention requirements.
Designed for Microcontrollers, Not Linux Servers
The Koios telemetry and logging agent is built for constrained devices. Minimal RAM footprint. Configurable reporting intervals. Batched uploads over the same mTLS channel your device already uses — no additional attack surface, no extra connection overhead.
It runs in kilobytes, not megabytes. Because on a device with 64KB of RAM, every byte of your monitoring agent is a byte your application can't use.
Capabilities
Did Memory Usage Increase After v2.3.1?
Correlate metric changes with firmware deployments. Cross-reference log spikes with certificate rotations. See cause and effect across your fleet.
Koios automatically associates telemetry and log data with device identity, firmware version, and device group. Spot trends before they become incidents. Identify a memory leak affecting 3% of your fleet before a single customer reports an issue.