Pre-Deployment Validation

The Challenge

You've built new features or fixed bugs. Before deploying to production:

• Does it still work?
• Is performance acceptable?
• Does it work from all regions?

The Solution

Run Agent Status validation on your staging environment before production deployment.

Workflow

┌─────────────┐    ┌──────────────┐    ┌────────────┐    ┌──────────────┐
│   Develop   │ →  │  Deploy to   │ →  │  Agent     │ →  │  Deploy to   │
│   Feature   │    │   Staging    │    │  Status    │    │  Production  │
└─────────────┘    └──────────────┘    └────────────┘    └──────────────┘
                                              │
                                              ↓
                                        ┌──────────┐
                                        │  Failed? │ → Fix and retry
                                        └──────────┘

Setup

Register Staging Agent

curl -X POST "https://api.agentstatus.dev/api/v1/agents" \
  -H "Authorization: Bearer $AGENTSTATUS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "ChatBot - Staging",
    "endpoint_url": "https://staging.your-agent.com/v1/chat/completions",
    "schedule_type": "manual",
    "geos": ["us", "eu", "ap"],
    "max_nodes_per_run": 10
  }'

Note: schedule_type: "manual" — no continuous monitoring, only on-demand.

Validation Criteria

Minimum Passing Criteria

Metric	Threshold	Action
Verdict	UP	Required to proceed
Verdict	DEGRADED	Review before proceeding
Verdict	DOWN	Block deployment

Extended Criteria

Metric	Threshold	Action
Eval Pass Rate	≥ 95%	Required
P95 Latency	≤ 5000ms	Required
All regions UP	Yes	Required

Compare to Production Baseline

Before promoting, compare staging metrics to production:

• If staging is DOWN → block
• If latency regressed >50% → block
• If eval pass rate <90% → block

Canary Deployments

Deploy to 5% of traffic

Point Agent Status at canary endpoint

Validate for 15-30 minutes

If healthy, proceed with rollout

If degraded, rollback canary

Feature Flag Validation

Enable flag in staging

Run Agent Status validation

Compare to baseline (flag off)

If acceptable, enable in production

Best Practices

Mirror production config — Staging agent should have same prompts, regions

Use realistic data — Staging should reflect production patterns

Validate under load — Run load tests alongside Agent Status validation

Document baselines — Know what "normal" looks like

Automate everything — No manual steps in the happy path