NERVE
ekkOS can detect, diagnose, and heal problems in your codebase automatically.
NERVE is the self-healing layer of ekkOS. It monitors your memory system's health, detects anomalies like failing patterns or degraded performance, and takes corrective action — from quarantining bad patterns to rolling back broken deployments. Think of it as an immune system for your codebase.
How It Works
NERVE follows a pipeline — each anomaly flows through a series of stages before any action is taken. Nothing happens without evidence.
Continuously monitors your pattern health, success rates, and system metrics
Identifies anomalies — declining pattern success, production errors, latency spikes
Gathers evidence and produces a root-cause analysis with confidence scoring
Executes corrective actions — quarantine bad patterns, forge defensive ones, or roll back deployments
Checks that the fix worked. If it didn't, the action is reverted and logged
Autonomy Tiers
NERVE has four levels of autonomy. You control how much it can do on its own. Each tier unlocks more powerful actions, with stronger safety guarantees.
Read-only. NERVE detects and diagnoses problems but takes no action. You review everything.
NERVE can modify your memory — quarantine failing patterns, retire stale ones, forge defensive patterns from failures.
NERVE can generate pull requests to fix localized bugs. All changes go through PR review — nothing is auto-merged.
NERVE can automatically roll back a deployment if it detects an acute post-deploy outage. This is the fastest response — no human in the loop.
Dashboard Panels
The NERVE dashboard gives you real-time visibility into what the system is doing. Everything is observable — no black boxes.
Every anomaly NERVE detects gets a diagnosis with root-cause analysis, confidence score, and recommended actions.
Monitors the fast-path rollback system. Shows when reflexes evaluated, fired, or were blocked by safety guards.
Daily and weekly caps on autonomous actions. Prevents runaway healing — the system can't exceed its budget.
Overnight, NERVE recombines patterns from your library to discover cross-domain connections you might have missed.
Defensive patterns forged from past failures. When NERVE diagnoses a problem, it creates an antibody to prevent recurrence.
When the same anomaly triggers repeatedly, NERVE surfaces it as a goal — showing what needs structural improvement.
Safety Guarantees
NERVE is designed with multiple layers of safety. Every action is bounded, auditable, and reversible.
- Mutation budgets — Daily caps on every action type. The system cannot exceed its budget, even if multiple anomalies fire simultaneously.
- Confidence thresholds — Actions only execute when diagnosis confidence exceeds the tier's threshold. Low-confidence diagnoses are logged for manual review.
- Single-hop rollbacks — The Reflex Arc only reverts to the immediately previous deployment. No cascading reversions.
- Previous deployment health check — Rollbacks only fire if the previous deployment was confirmed healthy. NERVE won't revert to a known-bad state.
- Full audit trail — Every decision — including decisions not to act — is recorded with evidence, predicates, and safety check results.
- Kill switch — One click disables all autonomous actions immediately. The system goes read-only.
Vitals Integration
Connect your observability providers so NERVE can feel production pain — not just memory health.
Configure vitals in Dashboard → Settings → Vitals. Without vitals, NERVE monitors memory and session health signals only.
Getting Started
Open NERVE
Go to Dashboard → NERVE to see your system's health.
Set your tier
Start at R0 (observe only). Upgrade to R1 when you're comfortable with autonomous memory healing.
Connect vitals
Add your observability provider so NERVE can detect production issues, not just memory drift.