31 verified tools. 0 wrong bindings.
The record is measured, not asserted, across three real, unmodified applications, on every run. Every number on this page links to the dated experiment ledger in the repository that produced it.
Three real, unmodified applications.
| Application | Scale | Capabilities | Wrong bindings | Ledger |
|---|---|---|---|---|
| ERP backendServices return entities; discovers well. | 25 entities | 25 (16 by-id, 9 list) | 0 | ledger (opens in a new tab) |
| PMS backendServices return view objects; most reads refused by name. | 43 entities | 2 (2 by-id, 0 list) | 0 | ledger (opens in a new tab) |
| A real multi-module enterprise Spring applicationReuses entity names across bounded contexts; emits valid output where simple-name keying could not. | 169 modules / 111 entities | 4 | 0 | redacted |
Honestly. Coverage is idiom-bound; the guarantee is zero wrong bindings.
The tool discovers well when a service returns entities and poorly when it returns view objects, so coverage depends on the application's idioms, not on effort. The ERP app discovers 25 of its 25 entities; the view-heavy PMS app discovers 2 of 43. Two of the three apps share an author, so the counts do not compound. Your number depends on your idioms and could land anywhere in that range.
The one guarantee that transfers is the last column: zero wrong bindings, on every run of every phase, on all three applications. Preview your own app to see your real number before you decide.
What mis-bound 0 means, exactly: every generated tool is bound to the method the source proves backs it, so a tool never silently calls the wrong method or the wrong entity's data. It is a correctness guarantee about what each tool does.
It is not an access-control guarantee. Fail-loud does not authenticate callers or authorize actions; gating the endpoint is the operator's job (see security).
The third row is a real multi-module enterprise Spring application, kept private by agreement and shown by counts and shapes only, never by name, domain, or industry. It is the app that Entity Identity was built for: it emits valid output that simple-name keying could not produce, at 4 verified capabilities with zero wrong bindings.
The two public benchmarks link to their full per-entity ledgers above: every discovered capability with its binding, and every refusal with its located reason. Nothing here is summarized past what those dated records show.
ERP ledger (opens in a new tab)PMS ledger (opens in a new tab)Preview your own app