Ship changes without breaking production.

Every rule goes through a structured promotion pipeline. Shadow, canary, active - with instant rollback at every stage.

The Problem

Changing autonomous behavior is terrifying.

Changing autonomous behavior in production is terrifying. A new rule might fire when it shouldn't. A modified threshold might cascade into unexpected actions. Teams freeze, preferring known-bad behavior over unknown-different behavior.

The Lifecycle

A structured promotion pipeline

Every rule in Memrail goes through a structured promotion pipeline. New rules start silent, observing what they would do without executing. Then they execute for a small percentage of traffic. Only after validation do they go fully live.

1

Draft

Rules are authored and reviewed. Not yet evaluating against any traffic.

2

Shadow

Evaluates against real traffic but never executes. Logs what would have happened. Compare against current production behavior.

3

Canary

Executes on a small percentage of traffic. Sticky routing ensures the same user gets a consistent experience. Metrics compared against existing rules.

4

Active

Fully live in production. Validated through shadow and canary stages. Instant rollback available at any time.

Rule lifecycle showing draft, shadow, canary, and active stages

For Teams

What this means for your team

Shadow mode

"Let me see what this rule would do on real traffic before I turn it on." Validate against production data without any risk of execution.

Canary mode

"Let me try this on 10% of traffic and compare outcomes." Gradual rollout with real metrics and real comparisons.

Instant rollback

"Something unexpected? One command to deactivate." No deployments, no waiting, no cascading side effects. Instant reversion to previous behavior.

Promotion gates

Define criteria that must be met before a rule advances to the next stage. Quality gates that enforce rigor without slowing teams down.

Autonomous behavior becomes a deployable, versionable, testable artifact - not a prayer.

See the rollout playbook in action

In the 14-Day Pilot, we demonstrate shadow mode validation on your workflow - real traffic, zero risk, complete visibility into what would change.

Start a Pilot

Questions

Frequently asked

How long should a rule stay in shadow mode?

It depends on your traffic volume and the rule's criticality. High-traffic workflows can validate in hours. Low-traffic or high-stakes rules might stay in shadow for days. The key is having enough data to compare shadow results against current production behavior with statistical confidence.

What happens if a canary rule causes unexpected behavior?

Rollback is instant. The rule returns to shadow mode or draft state immediately. Because canary rules only affect a small percentage of traffic with sticky routing, the blast radius is contained by design.

Can I run multiple canary rules simultaneously?

Yes. Each canary rule operates independently with its own traffic allocation and metrics. The system tracks interactions between concurrent canaries so you can measure combined effects and identify conflicts.