Ship dark. Roll out slow.
Toggle without deploying.
Turn features on and off and roll them out gradually — evaluated stably per user, drivable by API or your AI agent. The flat-rate, EU alternative to LaunchDarkly.
Feature flags shouldn’t cost more than your servers
LaunchDarkly and friends charge per seat and per monthly active user, live apart from your infra, and put an SDK plus a network round-trip on the hot path. koigrid gives you the essentials — instant toggle, gradual rollout, stable per-user evaluation — at a flat rate, right next to the apps they gate, evaluated with a single API call your agent can make.
How koigrid compares
| koigrid | LaunchDarkly | |
|---|---|---|
| Pricing | Flat rate | Per seat + per MAU |
| On/off toggle | Yes | Yes |
| Gradual % rollout | Yes (stable per user) | Yes |
| Lives next to your apps/DBs | Yes (same platform) | No |
| Evaluate by API / AI agent | Native | SDK |
| Attribute targeting | Roadmap | Yes |
| EU data residency | Yes | Optional |
From public docs, 2026. koigrid evaluates flags server-side with a deterministic hash, so a given user always sees the same variant.
Everything a rollout needs
Toggle without deploying
Flip a feature on or off instantly — no redeploy, no restart.
Gradual rollout
Ramp from 1% to 100% by percentage. The assignment is stable per user, so nobody sees the feature flicker between requests.
Stable per-user evaluation
A deterministic hash of flag + context puts each user consistently in or out — reproducible and fair.
Agent-native
Create, toggle and roll out flags from Claude Code or any LLM via the API — full parity with the dashboard.
One evaluate call
POST /flags/evaluate with a user id returns every flag as a boolean — evaluate once per request and cache.
Project-scoped
Flags live inside a project, right next to the app, database and cache they gate.