Cognition
True cognition in production
Cognition does not just collect data. It understands your Node.js systems, connecting errors, runtime health, and security signals into actionable intelligence that can act, fix, and defend with a human in the loop.
npm i @skytells/cognitionHEAP USED
256MB
EVENT LOOP
4.2ms
CPU
12.4%
ERRORS
0
AI cognition streaming
0 threats · 128 events analyzed · baseline stable
Your systems deserve intelligence, not just monitoring.
One zero-dependency SDK turns raw production signals into context, root cause, and controlled action. Every capability contributes to one continuously correlated view of production.
Automatic error capture
Full execution context, automatically.
Every uncaught exception and unhandled rejection is captured with full stack traces, breadcrumbs, and the context needed to resolve it quickly.
Runtime observer
Continuous process health in real time.
Track event-loop lag, memory and CPU usage, garbage-collection pressure, active handles, and leaks with configurable thresholds.
Security scanner
Threat detection at the request boundary.
Scan incoming requests for SQL injection, XSS, path traversal, and command injection through Express middleware or a standalone function.
OpenTelemetry bridge
Existing traces, one transport.
Enable the optional OpenTelemetry bridge to ship spans through the Cognition transport without rebuilding your tracing setup.
Zero dependencies
Node.js built-ins, ESM, CJS, and TypeScript.
The production SDK uses Node.js built-ins only and ships dual ESM and CJS formats with complete TypeScript declarations.
Event filtering and PII
Control every event before it leaves.
Use beforeSend to redact, transform, or drop telemetry. Your application decides exactly which information reaches Cognition.
Breadcrumbs
The trail leading to every failure.
Console output is captured as breadcrumbs, and custom breadcrumbs can include a category, message, metadata, and severity.
AI-driven analysis
Root cause, response, and controlled action.
AI correlates errors, runtime conditions, and threats into a timeline, then prepares actions to fix and defend with human approval where required.
Three layers working together.
Cognition watches the process from three angles, then correlates every signal into one timeline instead of leaving you with isolated events.
01Error captureSIGNAL / error.unhandled+
Uncaught exceptions and unhandled rejections arrive with stack traces, context, and breadcrumbs. Nothing slips through without an explanation trail.
02Runtime observerSIGNAL / runtime.snapshot+
Memory, CPU, event-loop utilization, garbage collection, and active handles show what the process was doing when behavior changed.
03Security scannerSIGNAL / security.threat+
Request inputs are checked for injection and traversal patterns before they escalate into a production incident.
Up and running in under five minutes.
01Create a project
Create a Skytells project, give it a name, and keep the Project ID ready for initialization.
Create project02Get your API key
Create a server-side key from API Keys in project settings and keep it private in your environment.
Open API Keys03Install and initialize
Install the SDK and call Cognition.init() once in your Node.js entry point. Capture starts immediately.
Read setup docs1import { Cognition } from "@skytells/cognition";23Cognition.init(({)4 apiKey: process.env.SKYTELLS_API_KEY!,5 projectId: process.env.SKYTELLS_PROJECT_ID!,6});78// AI-powered cognition is live.Know what happened, why it happened, and what to do next.
Cognition connects errors to runtime conditions, maps threats to targeted request paths, and prepares a root-cause timeline, response, and defense playbook with human approval where needed.
When the Gulf region went dark, Cognition kept the lights on.
During the Gulf Region Infrastructure Disruption in March 2026, Cognition detected abnormal behavior, correlated the failure domain, and informed the recovery path that restored full service in under 60 minutes with zero data loss.
T+0 minAnomaly detected
Runtime signals identified elevated error rates and event-loop latency across affected services.
T+10 minRoot cause correlated
AI analysis connected infrastructure telemetry with error patterns and isolated the regional failure domain.
T+15 minAutomated failover triggered
Actionable intelligence fed the multi-vendor recovery strategy and traffic moved to alternate capacity.
T+55 minFull service restored
Services returned to baseline while Cognition continued verifying stability across restored nodes.
<60 min
Full recovery
0
Data loss
5 min
Detection
100%
Continuity