ConfigCat is the closest thing to wagle that already shipped: a hosted service whose SDKs evaluate locally, priced on config fetches rather than seats, and deliberately kept simple. It also does several things wagle can't touch today: about 21 open-source SDKs including mobile, every feature on every tier (SSO/SAML/SCIM even on the free plan, which is genuinely rare), eight years and 500+ customers behind it, an SLA, and a paid on-prem option. And it is itself bootstrapped and independent, so wagle's usual "no acquisition risk" card does not apply here: ConfigCat has the same structural independence with a real track record on top. Do not pick wagle if you need broad or mobile SDK coverage, SSO/SAML/SCIM today, an on-prem deployment, or a proven vendor. Consider wagle only if you specifically want versioned conditional sync (every poll sends the held version and an unchanged answer is a tiny response, so short intervals stay cheap, versus ConfigCat's fixed CDN polling where each check bills a config download), native JSON typed config with atomic per-version snapshots, and AI agents as revocable scoped principals rather than shared API keys, accepting, in exchange, a young product run by one person. Neither of us runs experiments, by design: a flag service flips flags and ships config, and you measure outcomes with your own analytics stack.
Competitor claims re-verified 2026-07-19 · sources at the bottom
In your process. SDKs seed a snapshot and evaluate locally, byte-identical across the Go SDK, TS SDK and server, pinned by shared CI vectors. Same hosted-plus-local-eval family as ConfigCat; this row is a tie on model, and ConfigCat wins on the breadth of it.
Also local: each of ~21 SDKs downloads the config JSON and evaluates targeting rules in-process, the same architectural family as wagle, shipped and proven across far more languages.
A versioned conditional poll: the SDK sends its held version and an unchanged check is answered with a tiny "nothing new" response, so tight intervals stay cheap. The interval is per client, the freshness-vs-cost dial.
Strict periodic polling of the config CDN: 60s by default, tunable down to 1s at the cost of more billed downloads. There is no server push, SSE or streaming, so a change is invisible until the next poll. In exchange the model is CDN-cached, dead simple to operate, and keeps serving on outage: the same pull-based family as wagle, with a fixed cadence and a billed download per check.
Two languages: Go and TypeScript/React. No Python, Java, .NET, Ruby, Rust, and no mobile. If your stack needs those, ConfigCat is the honest answer.
About 21 open-source (MIT) SDKs across backend, web, and mobile: .NET, Java, Android, Swift/iOS, Kotlin, Flutter, Go, Python, PHP, Ruby, Rust, Unity, Unreal and more.
Native BOOL / NUMBER / STRING / JSON values, one atomic write path with per-project versioning, and history diffs down to the inner JSON key. Native JSON config is the modest edge here.
Four scalar setting types: on/off, text, whole number, decimal. JSON is stored as a text setting and managed by hand; the common scalar cases are well covered.
Three fixed per-project roles (admin / editor / viewer). No SSO/SAML/SCIM, no custom roles: if your org lives in Okta, that gap is real today.
SSO, SAML, 2FA and SCIM on every tier, including the free plan, plus permission groups. Shipping enterprise auth to free accounts is rare, and it is a straightforward win over both wagle and the tools that gate it behind Enterprise.
Agents are first-class principals: project-scoped 24h tokens, live revocation, and MCP that the server enforces as agent-only. The surface is narrower, but the agent is a revocable, scoped identity rather than a shared key.
Ships a broader read-write MCP server: products, configs, environments, flags, segments, webhooks, stale-flag detection, code references. It authenticates with a long-lived Management API key, not a per-agent principal.
None: wagle is hosted-only, by design. If on-prem or data residency is a requirement, wagle is the wrong tool and says so.
A full on-premise Docker deployment exists (dashboard, CDN, API), but the management platform is proprietary and the on-prem offer is custom, sales-quoted pricing. For an open self-host instead, look at Unleash or Flagsmith.
One meter: SDK syncs, with local evaluations free and unlimited. The same charge-for-the-pull philosophy as ConfigCat, on one number instead of tiered download volumes.
Flat, seat-free tiers metered on config downloads, with a genuinely generous free plan (10 flags, 2 environments, 5M downloads/month, SSO included). The same charge-for-the-fetch philosophy as wagle, with tier limits on flags and environments.
Third-party sources only: reviews, buyer data, and their own published docs and pricing; never their marketing copy, never our opinion of them.
The why & how page states every deliberate limitation plainly, the fastest way to rule wagle in or out.