A boolean flag in a config file is trivial. Evaluating it reliably in milliseconds for millions of requests, with instant global updates and gradual rollout percentages, is genuine distributed-systems infrastructure — the same reasoning that makes LaunchDarkly a JUST PAY here.
Just pay
Effort: not worth startingRunning cost: $0/mo to start
Why CanICodeThis says Just pay
A simple on/off flag checked from an environment variable or a database row is genuinely basic
For a small personal project with a handful of flags, checking a config value directly is a real, achievable approach
Toggling a flag manually is straightforward
The parts that'll cause problems
Low-latency flag evaluation at real production traffic, with instant propagation when a flag changes, is a genuine distributed-systems problem
Gradual percentage-based rollouts with consistent user bucketing need careful, correct hashing logic
Audit logs and approval workflows for flag changes at company scale are real governance features
How I'd build this
frontend
Next.js
database
Supabase
hosting
Vercel
MVP scope
If you proceed: flags checked directly from a config table in your own database
Manual on/off toggling
Postpone to v2
Distributed low-latency evaluation at scale (genuinely hard — consider an existing service instead)
Gradual percentage rollouts with consistent bucketing
What it actually costs to build
AI agent — with a subscription (Claude Pro/Max, Cursor, etc.)
$0 marginal
AI agent — pay-per-use API, no subscription
$86–$172 one-time
Hosting, once it's running
$0/mo (free tier)
Domain name, if you want your own
~$12/yr
The build prompt
Don't build a distributed feature-flag service — check flags directly from your own config, genuinely sufficient at personal-project scale:
1. A flags table in your own database (name, boolean value, optionally a percentage for gradual rollout).
2. Check the flag value directly in your code where you'd otherwise hard-code behavior — a single database read, cached in memory if you want to avoid a query per request.
3. A simple admin page to toggle flags manually.
That's real, working feature flagging for a small project. Out of scope: low-latency flag evaluation at real production traffic with instant global propagation when a flag changes — a genuine distributed-systems problem, precisely what a dedicated service exists to solve — and consistent user bucketing for percentage rollouts, which needs careful, correct hashing logic to get right.