Causal RCA
Graph-backed traversal across services, deploys, schemas.
THE PLATFORM
Autonomous engineering, supervised by you.
NEXIS is a closed-loop incident-response platform built for SRE and platform teams. It detects anomalies, traces root cause, generates and validates patches, and routes them through engineer approval — so your team only sees the diffs that need a human.
A live recovery, mid-flight.
When Sentinel flags an anomaly, every agent hand-off lands in the audit log within seconds. This is what you see on the live pipeline canvas in /console/incidents.
incident-2847.log
[00:02:14] sentinel › anomaly detected — pipeline: etl_orders▍
DETECTION
Sentinel — anomalies caught before customers feel them.
Sentinel ingests signals from Sentry, Datadog, OTel collectors, GitHub Actions runs, and your own metrics endpoints. It correlates them into a single anomaly stream and routes high-confidence events directly into the recovery pipeline.
- Streaming anomaly model — flags events within 90 seconds of occurrence.
- Severity routing: P0 → on-call DM, P1 → channel, P2 → ticket queue.
- False-positive suppression learned per workspace, never global.
- Replays the last 24h on every new integration so you start with context.
metric 1
94%
metric 2
1.2s
metric 3
0.3%
metric 4
12k
Live preview is illustrative — real telemetry varies per workspace.
RECOVERY
Pathfinder + Synthesiser — root cause and a working patch.
Pathfinder builds a causal graph across services, deploys, schemas, and recent migrations. Synthesiser uses that graph plus your contract repo to generate candidate patches — then QA writes property tests to validate them in a Docker-isolated shadow pipeline.
- Neo4j-backed causal traversal — explains why, not just what.
- Patch candidates ranked by blast radius, test coverage, and rollback cost.
- Shadow pipeline mirrors prod topology with zero customer traffic.
- Every candidate carries a plain-English summary for the approver.
metric 1
2,847
metric 2
0.18s
metric 3
0.4%
metric 4
12
Live preview is illustrative — real telemetry varies per workspace.
APPROVAL
Approval Gate — engineers stay in the loop on what matters.
Severity-routed approvals reach the right human via Slack, PagerDuty, or email. The approver sees the diff, the failing-then-passing test, and a one-paragraph summary. Approve and the fix ships through your existing deploy tooling (ArgoCD, GitHub Actions, custom).
- One-click approval with cryptographic audit trail.
- Approval policies per project — auto-merge for cosmetic, human for schema.
- Auto-rollback if post-deploy health checks fail within 10 minutes.
- Mobile-friendly approval surface; median engineer time 60 seconds.
metric 1
60s
metric 2
1-click
metric 3
100%
metric 4
0
Live preview is illustrative — real telemetry varies per workspace.
CAPABILITIES
Everything you'd expect from an incident-response platform.
And a few things you wouldn't — like a metered energy wallet and signed pull requests.
Patch synthesis
LLM patches constrained to your contracts and schemas.
Shadow validation
Docker-isolated pipeline mirrors prod topology.
Property testing
Auto-generated unit + property suites per candidate.
Severity routing
P0/P1/P2/P3 policies per project, no global defaults.
Audit trail
Typed agent hand-offs persisted forever, immutable.
Live pipeline canvas
Watch every stage of recovery in real time.
Energy wallet
Per-token + per-recovery metering, transparent costs.
Schema migrations
Data Engineer agent owns backfill + cutover plans.
Pull-request emission
Patches land as PRs on your repo, signed and verified.
Role-based access
Approval policies and budgets per team and per project.
Compliance-ready
SOC 2 Type I in progress; air-gapped option on Enterprise.
Ready to put the fleet to work?
Free tier ships with a synthetic incident generator so you can try the full pipeline without wiring up production.