FOR PLATFORM ENGINEERS

Add a governance gate to your pipeline
that actually knows production.

Legible integrates into your CI/CD pipeline as a deployment verification gate. Before a deployment promotes to production, Legible checks it against a stage-validated behavioral boundary built from real telemetry — not just tests or policies.

OpenTelemetry nativeWebhook-drivenRead-only telemetryZero SDK< 5 min setup

Request Access

Architecture

Where Legible sits in your stack

Not a replacement for anything — the missing layer between your observability platform and your deployment pipeline.

🚀

Your CI/CD

GitHub Actions, GitLab CI, ArgoCD, Jenkins

→

🛡️

Legible Gate

Verification API before promotion

→

🏭

Production

Your services in production

Inbound — Legible reads

●OpenTelemetry traces via OTLP gRPC/HTTP

●CI/CD webhooks (deploy events, PR metadata)

●Feature flag change events

●Configuration change notifications

Outbound — Legible provides

●Verification verdict + evidence chain

●Deployment control recommendation

●Governance health metrics

●Boundary composition breakdown

Pipeline Integration

Add the gate in one pipeline step

A single API call in your deployment pipeline. Legible returns a verdict with full evidence chain.

GitHub Actions workflow — deploy.yml
# .github/workflows/deploy.yml
jobs:
  deploy:
    steps:
      - name: Deploy to staging
        run: helm upgrade payment-svc ./charts/payment-svc -n staging

      - name: Legible verification gate
        run: |
          VERDICT=$(curl -s -X POST https://api.legible.systems/v1/verify             -H "Authorization: Bearer ${{ secrets.LEGIBLE_TOKEN }}"             -H "Content-Type: application/json"             -d '{
              "change_event_id": "${{ github.run_id }}",
              "service": "payment-svc",
              "version": "${{ github.sha }}",
              "environment": "staging"
            }')

          RECOMMENDATION=$(echo $VERDICT | jq -r '.recommendation')

          if [ "$RECOMMENDATION" = "HOLD" ] || [ "$RECOMMENDATION" = "ESCALATE" ]; then
            echo "::error::Deployment blocked by Legible: $RECOMMENDATION"
            exit 1
          fi

      - name: Promote to production
        if: success()
        run: helm upgrade payment-svc ./charts/payment-svc -n production
      

Verification response — ALLOW

{
  "recommendation": "ALLOW",
  "verdict": "MATCHED_INTENDED",
  "confidence": "HIGH",
  "evidence_summary": {
    "stage_validation_verdict": "CONFIRMED",
    "explained_changes": 3,
    "unexplained_changes": 0,
    "invariant_violations": 0
  },
  "auto_promote_eligible": true
}
      

Configuration

Governance rules as code

Define invariants, thresholds, and governance policies per workflow. Version-controlled, auditable, reviewable.

legible.governance.yaml
workflow: checkout-flow
environment: production

invariants:
  required_nodes:
    - fraud-check
    - payment-authorization
    - inventory-reservation
  forbidden_transitions:
    - from: api-gateway
      to: bank-integration    # must go through payment-svcretry_ceiling: 5
  latency_hard_max_ms: 3000

thresholds:
  structural_conformance_min: 0.7
  distributional_max_shift: 0.20
  retry_max_amplification: 2.5

stage_validation:
  sbd_wait_timeout_minutes: 30
  auto_promote_on_confirmed: true
      

Data Model

What Legible stores

Versioned, immutable artifacts with full provenance. Every governance decision traces back to raw telemetry.

📊

Workflow Baselines

Versioned graph structure, path frequency distributions, invariant rules, capability scores. Immutable once promoted.

📦

Change Envelopes

Per-deployment predictive envelopes with confidence, assumptions, evidence provenance, and TTL.

🔒

Trusted Change Boundaries

Stage-validated behavioral surface. Union of SBD + confirmed IIC + transitive deps + divergences.

🧪

Stage Behavioral Deltas

Multi-dimensional measurement of stage behavior: structural, distributional, retry, temporal.

📋

Delta Conformance Records

Production vs boundary comparison. Per-dimension scores, explained/unexplained breakdown, verdict.

⚖️

Classifications

Final verdict with evidence chain, human override status, governance recommendation, audit trail.

API Surface

REST API for everything

Every artifact, metric, and governance decision is available via API.

POST/v1/verifySubmit deployment for verification

GET/v1/verifications/{'{id}'}Verdict + evidence + boundary

GET/v1/workflows/{'{id}'}/baselineCurrent baseline for workflow

GET/v1/governance/healthGovernance health dashboard

GET/v1/governance/anti-gamingAnti-gaming signals

POST/v1/feedbackSubmit human review verdict

GET/v1/workflows/{'{id}'}/divergencesDivergence model entries

GET/v1/workflows/{'{id}'}/boundary-historyBoundary composition history