Built for teams before Series A. Scalable anyways after it.

Ship like a big company. Pay like a small one.

Release control built for startups moving fast — safe rollouts, no redeploys, and pricing that scales only when you do.

Production-ready from day one. Built on modern edge architecture for low-latency evaluation.
~25ms

global evaluation

0

SDK lock-in

300+

edge locations

$0

to start

No contracts. No surprise jumps. Upgrade when it makes sense.

What this unlocks for your startup

The release workflow you need to move fast without breaking trust.

  • Ship unfinished work without exposing it: merge to main behind a flag — users never see it
  • Roll out to 1% before you roll out to everyone: percentage targeting, segment rules, your pace
  • Turn off a bad release in seconds: flip a flag, no redeploy, no hotfix
  • Test with real users before you commit: validate features and pricing on actual traffic

Start simple. Stay in control.

Three steps from zero to controlled rollout.

1

Create a flag

Define what you want to control. Name it, set a default, done.

2

Evaluate via HTTP

One request returns every flag for a given user. No polling, no sync.

GET eval.onlyflags.app/{project}/{context}
3

Roll out gradually

Target by segment, percentage, or context. Roll back without a redeploy if something breaks.

Works with any stack

One HTTP call. No SDK required. Pick your language and ship.

See code examples →

Built for teams before Series A

Enterprise flag platforms optimize for governance.
We optimize for shipping.

  • No contracts, no calls: start in minutes
  • No heavy governance: ship without process tax
  • No SDK lock-in: HTTP-first, runs anywhere
  • Costs stay tiny pre-traction: scale only when usage grows

Fast everywhere. Not just on paper.

Evaluated at the edge, closest to your users.

Your flags load before your UI does

No loading spinners. No layout shifts. Your users see the right experience fast — whether they're in Tokyo, São Paulo, or Paris.

Half of all requests resolve in under 25ms. Even the slowest 1% take less than 111ms — faster than a blink. Release control should never be the reason your app feels slow.

~25ms

global evaluation latency

<111ms

fast toggles without SDK overhead

300+

edge locations for multi-region delivery

Evaluated close to your users on globally distributed infrastructure.

No enterprise tax

You shouldn't need a sales call to toggle a feature.

Setup
Complex SDK setup
One HTTP call, any language
Pricing
Expensive from day one
Free trial, then $0.03 / 10K evals
Getting started
Sales calls & contracts
Start in minutes
Long-term
"Starter tools" you outgrow
Start small. Never migrate.

Pay like a small team until you're not one — without replatforming.

Everything you need to ship safely

No feature gates. No per-seat pricing. One subscription, full platform.

Edge-evaluated

Your flags resolve before your UI paints. 300+ edge locations, ~25ms median response — your users never wait.

No SDK required

One HTTP call from any language. No dependency, no version conflicts, no lock-in. You own the integration.

Safe rollouts

Release to 1%, then 10%, then everyone. Something breaks? Turn it off without redeploying.

Multi-variant flags

Run A/B/C experiments with percentage-based traffic splits. Not just on/off — real experimentation from day one.

19 targeting operators

Target by plan, country, version, date, regex, semver, and more. Powerful enough for complex rules, simple enough to set up in minutes.

Unlimited everything

Flags, seats, and projects are unlimited — even during your free trial. No feature gates. No artificial limits.

AI-native release control

OnlyFlags exposes an MCP server. Your LLM can create flags, configure targeting rules, and manage segments directly.

Add to Claude Code, Codex, or Cursor

Create a user token in your dashboard and paste this into your MCP client config.

{
  "mcpServers": {
    "onlyflags": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://api.onlyflags.app/api/mcp",
        "--header",
        "Authorization:${AUTH_HEADER}"
      ],
      "env": {
        "AUTH_HEADER": "Bearer of_your_api_key_here"
      }
    }
  }
}

What your LLM can do

  • List workspaces — one key, all your workspaces
  • Create, list, update, and delete flags
  • Create, list, update, and delete segments
  • Full targeting rules and strategy configuration

Start small. Never migrate.

Use the same release workflow from your MVP to your growth stage. The pricing and infrastructure scale with you — without forcing a platform switch later.

MVP

One flag, one rollout

Growth

Targeting + percent rollout

Scale

Global performance + reliability

Built by developers, for fast teams

I never wanted to spend money on tooling — not even $20 — before getting my first user. So I built something I'd actually use: release control that costs nothing until it needs to, and scales without switching platforms.

Same release infrastructure from MVP to real scale.

No contracts. No tiers. Just usage-based pricing.

Pay like a small team until you're not one.

One flat rate. Costs scale linearly with your usage — nothing else.

No contracts No per-seat fees Pay only for what you use

Usage-based

$0.03 / 10K evaluations

7-day free trial, no credit card required

  • Unlimited flags
  • Unlimited seats
  • Unlimited projects
Start free

What it costs at scale

Monthly evaluations Cost
10K $0.03
100K $0.30
1M $3
10M $30
100K evals/month
$0.30
$10–45/mo + per-seat fees
1M evals/month
$3
$45–325/mo + per-seat fees
10M evals/month
$30
$100–500/mo + per-seat fees

Based on published pricing from major release control providers as of 2026.

Frequently asked questions

Real answers to the questions developers actually ask.

How does it work without an SDK?

OnlyFlags evaluates flags at the edge. You send a single GET request with your project key and a base64-encoded context object. The response is a JSON object with all your flag values. We provide ready-to-use code snippets you can copy, paste, and modify — you own the code entirely. No dependency to manage, no version conflicts, and it works from any language or platform that can make HTTP requests.

What happens if the service goes down?

Flag evaluations run across 300+ edge locations with built-in redundancy. Even during the unlikely event of an origin issue, your users continue receiving flag values. We recommend using sensible defaults in your code as a best practice.

What happens after the free trial?

After the 7-day trial, you move to usage-based billing at $0.03 per 10K evaluations. No commitment, no credit card upfront. You still get unlimited flags, seats, and projects — you only pay for the evaluations you actually use.

Will pricing spike suddenly?

No. Pricing is purely usage-based at a flat rate of $0.03 per 10K evaluations. No per-seat fees, no hidden charges, no tiers to jump between. Costs scale linearly with your usage — nothing else.

Will I need to migrate later?

No. OnlyFlags is built to scale with you. The same infrastructure, the same API, and the same workflow from your first flag to millions of evaluations. You never need to switch tools as you grow.

Can I use it from the frontend?

Yes. Since OnlyFlags is just an HTTP API, you can call it from anywhere — browser JavaScript, mobile apps, server-side code, edge functions, even IoT devices. The response is plain JSON, so no special client library is needed.

Where is my data stored?

Flag configurations are stored and replicated across a global edge network. Evaluation requests are processed close to your users and are not logged or stored — we never see your user data. The context object you send is used only for real-time evaluation and is discarded immediately.

Is this production-ready?

Yes. Flags are evaluated across 300+ edge locations worldwide with built-in redundancy. Median response time is ~25ms globally, and even the slowest 1% of requests resolve in under 111ms. The infrastructure is built for reliability from day one — not bolted on later.

Why don't flag changes apply instantly?

They do — in the same region. When you update a flag, the change is instant for requests hitting the same edge location. Propagation across all 300+ edge locations may take up to 1–2 minutes. For most use cases — rollouts, feature gates, A/B tests — this is invisible to users.

Start for free. Scale when you're ready.

No credit card. No startup tax. Start building in minutes.

Create your first flag