I built a release control room for feature flags — Compass Ultra

Feature flags were supposed to make releases safer.

Over time, they become a release surface of their own: stale flags, risky rollouts, missing owners, hidden dependencies, production overrides, and Slack threads pretending to be audit trails.

I built Compass Ultra to fix that. It's a release control room that wraps a structured review workflow around your existing feature flags — before production changes go live.

It is not a feature flag provider. It's the review layer that sits around the ones you already use.

The Problem

If your team ships behind feature flags, you know this feeling:

• 10 flags are "on" in production. Who owns each one?
• Did anyone check whether flash_sale_engine depends on payments.stripe_v4 before enabling both at 100% rollout on Black Friday eve?
• Your canary is at 60% but company policy says 50% max. Did anyone catch that?
• The CAB meeting is in an hour. Where's the runbook?

Most teams answer these questions through a mix of Slack threads, spreadsheets, and institutional memory. That works until it doesn't.

What Compass Ultra Does

Compass Ultra gives you a single workspace where you can:

1. Load your release state — import flags from LaunchDarkly, Statsig, Unleash, Flagsmith, or Firebase Remote Config (read-only, via a backend proxy so your keys never touch the browser)
2. Evaluate every flag against real user context — by user key, email, tenant, plan, role, region, country, device, and environment
3. Run 9 automated enterprise policy gates — traceability, approvers, expiry, canary limits, overrides, dependency health, and more
4. Get an AI risk analysis — ship / hold / fix-first verdict with specific blockers, affected flags, financial impact estimates, and remediation steps
5. Compare snapshots — see exactly what changed between two release checkpoints
6. Export a Release Readiness Certificate — a CAB-ready PDF with flag evaluations, policy results, risk summary, rollback notes, and approver sign-off list

The AI Risk Analysis (This Is the Part That Surprised Me)

I honestly didn't know how useful this would be until I ran it against the demo.

The live AI came back and flagged:

• 3 canary-required flags exceeding 50% rollout — a hard policy block
• promos.flash_sale_engine at 100% rollout depending on both payments.stripe_v4 (100%) and inventory.realtime_sync (60%) — a cascading dependency chain at full peak traffic exposure
• Multiple flags expiring within 4 days of the deploy window
• A 23:00–01:00 ET Thanksgiving night deploy with no canary phase planned

Every line of that is accurate and specific. It's not generic "you should test more" advice. It reads like a senior SRE walking you through a pre-deploy review.

There's also a deterministic fallback — if the live AI service is unavailable, a local rule-based analyzer runs automatically so you're never blocked.

The 9 Policy Gates

Every workspace state change runs these checks automatically:

Provider Integrations

Compass Ultra syncs flag state read-only from:

• LaunchDarkly
• Statsig
• Unleash
• Flagsmith
• Firebase Remote Config

Your API keys go into the backend proxy. The browser never sees them.

The CI Release Gate

Block deploys in CI when release risk exceeds a threshold:

- uses: ./.github/actions/compass-check
  with:
    compass_api_key: ${{ secrets.COMPASS_API_KEY }}
    risk_threshold: high

The Embeddable AI DevOps Widget

Drop a floating AI assistant onto any page with a single script tag:

<script src="https://www.compassultra.com/ai-devops-widget.js"></script>

It reads the live workspace state, answers release questions in natural language ("Is it safe to ship?", "What are the blockers?", "What's the rollback plan?"), and routes DevOps status questions through a backend search proxy.

Pricing

Free tier includes the full policy gate suite, flag evaluation engine, and PDF export — no card required.

Try It

Live app: https://www.compassultra.com
Demo (no login): https://www.compassultra.com/app?demo=true
AI DevOps checker: https://www.compassultra.com/ai-devops

The demo is pre-loaded with a simulated Black Friday retail release — 10 flags across LaunchDarkly, Statsig, and Firebase, with a policy blocker and two warnings already in place. Run the AI analysis and see what it flags.

Built for teams that ship fast and still need proof before production. Happy to answer questions or take brutal feedback in the comments.