Flagship
Deploy on Monday. Release on Thursday.

Flagship is the feature-flag and progressive-delivery platform for teams that ship constantly. Merge the code, dark-launch it behind a flag, serve it to 1% of traffic, watch the guardrails, then ramp to everyone — or kill it in 200ms — without touching the deploy.

  • Deploy ≠ release
  • 200ms global kill switch
  • Roll back without a redeploy
checkout.ts
const checkout = flagship.variant("new-checkout", {
  user: ctx.user,
  rules: [{ segment: "beta", serve: "on" }],
  rollout: { on: 5 },        // 5% of everyone else
  fallback: "off",           // safe if we go dark
});

if (checkout === "on") renderNewCheckout();

Release teams ramp every change through Flagship

CohortSwitchbackNorthsailRelayStratusHemlockBrightwireTollgateCohortSwitchbackNorthsailRelayStratusHemlockBrightwireTollgate
The control plane

Every release becomesa dial, not a leap.

Stop binding 'it's deployed' to 'it's live.' Flagship gives every team one place to gate, target, measure, and reverse any change already running in production.

Percentage rollouts

Move a release from 1% to 100% on a slider or a schedule. Sticky bucketing pins each user to one variant across sessions, devices, and replicas — no flicker between visits.

200ms kill switch

A bad release is one toggle from gone. The off state reaches every edge node and connected SDK in under 200 milliseconds, worldwide — no deploy in the loop.

Targeting rules

Serve variants by plan, region, account, app version, or any attribute you stream in. Compose rules in a no-code builder; every one is versioned and diffable underneath.

Automated guardrails

Bind error rate, latency, and conversion to each rollout. The moment a variant regresses past your threshold, Flagship halts the ramp and pages the owner — before the blast radius grows.

Full audit trail

Who flipped which flag, for whom, from what value, and when. Every change is logged, diffable, and reversible to any point in the timeline.

What progressive delivery looks like in production

11ms
p99 edge flag evaluation
200ms
Global kill-switch propagation
9B
Flag checks served daily
73%
Fewer rollback deploys
Built for the people on call

Flags that live inyour codebase,not a spreadsheet.

Typed SDKs, in-process evaluation, and a sandbox that mirrors production. Wrap a flag in one line and trust it under load.

Local evaluation

SDKs evaluate flags in-process from a ruleset streamed to memory. Zero network call on the hot path, zero added latency per request — sub-millisecond median.

Typed flag SDKs

First-class, fully typed SDKs for TypeScript, Go, Python, Rust, Swift, and Kotlin. Your editor knows every variant and its return type before you ship.

Flags as code

Define, review, and promote flags through pull requests with our Terraform provider and GitHub Action. Production state never drifts from main.

Stale-flag cleanup

Flagship scans your repo, finds the toggles you shipped and forgot, and opens the PR that removes them and their dead branch. Tech debt that retires itself.

One flag, four motions

From a quiet betato a global launch.

The same flag carries a change through its whole life. No second tool to learn at each stage.

Target

Dark-launch it internally

Serve the change to your own team and a named beta segment while everyone else stays on the old path. Real feedback before a single customer is in the blast radius.

Rollout

Canary the risk

Release to 1% of traffic with guardrails watching error rate and latency live. Catch the regression at 1% and auto-halt — never discover it at 100%.

Measure

Run the experiment

Split traffic across variants and read statistically sound lift on the metrics you actually report — a built-in sequential test calls the winner without you waiting on a fixed sample.

Ship

Ramp, then clean up

Promote to everyone, archive the flag, and let Flagship open the PR that deletes the dead code path so the win doesn't leave a mess behind.

From the on-call channel

The teams who stoppedbracing for deploys.

We went from two release windows a week to shipping forty times a day. Every risky change rides a flag, so 'deploy' stopped being the word that emptied the room at standup.

L
Lena Vasquez
Director of Platform Engineering, Switchback

A pricing bug hit 100% of EU traffic at 2am. One toggle, 200 milliseconds, gone — no escalation, no hotfix deploy, no incident review the next morning. That single save paid for the contract.

D
Dev Patel
Staff SRE, Northsail

Product runs the rollouts now without filing a ticket, and engineering trusts the guardrails to catch anything they ramp too fast. Same dashboard, two teams, zero standoff.

M
Mara Olsen
VP Product, Relay
Pricing

Priced per seat, not per flag.

Create as many flags as your codebase needs. We never charge you to be careful.

Developer

For solo builders and side projects.

$0/mo
  • Up to 3 members
  • Unlimited flags
  • 1 environment
  • Community SDKs
  • 7-day audit history
Most popular

Team

For product teams shipping daily.

$24/seat/mo
  • Unlimited members
  • Unlimited environments
  • Percentage rollouts + targeting
  • Experiments & guardrails
  • 90-day audit history
  • Slack & PagerDuty alerts

Enterprise

For platform teams at scale.

Custom
  • SSO, SCIM & RBAC
  • Private edge or self-hosted relay
  • Unlimited audit retention
  • 99.99% SLA
  • Named delivery engineer
  • Approval workflows & change windows

The questions every release team asks.

Will flag checks slow down my app?

No. The SDKs evaluate flags locally from a ruleset streamed into memory, so there's no network round-trip on your request path. Median in-process evaluation is sub-millisecond; edge evaluation is 11ms at p99.

What happens if Flagship is unreachable?

Every SDK caches the last ruleset it received and falls back to the default you set per flag. If our service vanishes entirely, your app keeps serving the variants you last shipped — flags fail to a known state, never to a blank one.

How is this different from a config toggle we built ourselves?

Homegrown toggles give you on and off. Flagship adds sticky percentage rollouts, attribute targeting, experiments with real statistics, automated guardrails, a full audit trail, and stale-flag cleanup — every part that turns painful once more than one team is flipping things.

Can non-engineers manage rollouts safely?

Yes. Product and ops drive flags from a no-code dashboard, while approval workflows, change windows, and RBAC keep production fenced off. Every change a non-engineer makes is still logged and reversible to the second.

Do we have to send you our user data?

No. Targeting can run entirely on attributes evaluated inside your own infrastructure with the local SDKs or a self-hosted relay. Enterprise plans add a private edge, so no end-user data ever crosses your perimeter.

Put your next releasebehind a flag.

Install an SDK, wrap one line of code, and ship your first controlled rollout this afternoon. No sales call to get started.