What is Sidekick?
Sidekick is a self-hosted, open-source feature flag engine built for teams that need fast, reliable flag evaluation without sending data to a third-party SaaS vendor.
The Problem
Most feature flag services evaluate flags on their servers — meaning every isEnabled() call is a network request. This introduces:
- Latency — even a fast remote call adds 5–50ms per flag check
- Reliability risk — if the flag service is down, your flag evaluation breaks
- Vendor lock-in — migrating away is painful and expensive
- Data privacy concerns — user attributes are sent to an external service
The Sidekick Approach
Sidekick takes a different approach: local evaluation.
- Your server runs the Sidekick control plane (a single Rust binary)
- Each SDK connects once via SSE and receives the full flag set
- All evaluation happens in-process, in memory — no network calls
- When flags change, the server pushes deltas via SSE instantly
The result: sub-microsecond flag evaluation with real-time updates and zero external dependencies at evaluation time.
How It Compares
| Feature | Sidekick | LaunchDarkly / Statsig |
|---|---|---|
| Evaluation location | In-process (local) | Remote HTTP |
| Evaluation latency | ~100ns | ~5–50ms |
| Self-hosted | Yes | No (or limited) |
| Open source | MIT | Closed source |
| Vendor lock-in | None | High |
| Pricing | Free | Usage-based |
What Sidekick Is Not
- Not a A/B testing platform — Sidekick focuses on feature flags and rollouts, not experiment analysis
- Not a managed service — you operate the server yourself
- Not a data warehouse — Sidekick does not store analytics or events (that is your job)
Who Is It For?
- Teams that want full control over their feature flag infrastructure
- Applications where flag evaluation latency matters (hot paths, mobile apps, real-time systems)
- Organizations with data residency requirements that prohibit sending user data to third parties
- Developers who prefer open source and want to understand and extend the system