Feature Flags Guide: Implementation, Monitoring & Best Practices (2026)

// npm install @openfeature/server-sdk @openfeature/launchdarkly-provider
// OR @openfeature/flagsmith-provider / @openfeature/unleash-provider

import { OpenFeature } from '@openfeature/server-sdk';
import { LaunchDarklyProvider } from '@openfeature/launchdarkly-provider';

// Initialize (once at startup)
await OpenFeature.setProviderAndWait(
  new LaunchDarklyProvider(process.env.LAUNCHDARKLY_SDK_KEY)
);

const client = OpenFeature.getClient('my-service');

// Evaluate a flag with context
const userContext = {
  targetingKey: user.id,
  email: user.email,
  plan: user.plan,
  country: user.country,
};

// Boolean flag (release/ops)
const showNewCheckout = await client.getBooleanValue(
  'new-checkout-flow',
  false,          // safe default = old behavior
  userContext
);

// String flag (A/B test variant)
const buttonColor = await client.getStringValue(
  'checkout-button-color',
  'blue',         // default/control variant
  userContext
);

// Number flag (percentage limit)
const maxItemsPerCart = await client.getNumberValue(
  'max-cart-items',
  10,
  userContext
);

// Object flag (feature configuration)
const rateLimit = await client.getObjectValue(
  'api-rate-limit-config',
  { requestsPerMinute: 60, burstAllowance: 10 },
  userContext
);

// Monitor flag evaluation errors
client.addHooks({
  error: (hookContext, error) => {
    metrics.increment('feature_flag.evaluation_error', {
      flag: hookContext.flagKey,
      error: error.message
    });
  }
});

// LaunchDarkly: percentage rollout via UI
// 1. Create "new-payment-processor" boolean flag
// 2. Add targeting rule: percentage rollout — 5% true, 95% false
// 3. Deploy code with flag check
// 4. Monitor metrics for 5% vs 95% cohorts
// 5. Increase to 20% → 50% → 100% if metrics look good

// In code: the flag handles cohort assignment automatically
const useNewPayment = await client.getBooleanValue(
  'new-payment-processor',
  false,
  { targetingKey: userId }
);

// Monitor canary health — track metrics per flag variant
const paymentResult = await (useNewPayment
  ? newPaymentProcessor.charge(order)
  : legacyPaymentProcessor.charge(order)
);

// Emit metrics with flag variant as dimension
metrics.increment('payment.charge', {
  success: paymentResult.success,
  processor: useNewPayment ? 'new' : 'legacy',
  error_code: paymentResult.errorCode
});

// Alert rules for canary monitoring:
// Error rate NEW vs LEGACY cohort
// rate(payment.charge{success=false,processor="new"}[5m])
//   > rate(payment.charge{success=false,processor="legacy"}[5m]) * 1.5

// Automatic rollback via feature flag API
async function autoRollback(flagKey: string) {
  if (errorRateRegression > 0.02) {
    await ldClient.updateFlag(flagKey, { percentage: 0 }); // instant rollback
    await alerting.page('on-call', 'Canary auto-rollback triggered');
  }
}