Facts, in tables, with the source named. Nothing here teaches; everything here answers.
Contracts
The two read most often are error codes and HTTP routes. What there is to compose is the capability catalog; what drives it is the CLI.
| Error codes | Every domain, every code, and what a client does with each |
| HTTP routes | Every route every capability mounts, with its verification strategy |
| Request context | Everything on c.var, and what it is when nothing filled it |
| Configuration schema | The root keys, and where each capability’s live |
Resources
| Bindings | Every binding name, what it points at, which are optional |
| Tables | Every pithy_* table and its owner |
| Naming rules and limits | One rule, ten namespaces, the character budgets |
| The config directory | File by file: path, permissions, what writes each |
The kit itself
| Package index | Every package, and which one is not MIT |
| Compatibility | The Node floor and the compatibility date |
| Supported locales | What ships, and how to add one |
| Accepted limits | The races that cannot be closed |
These pages are hand-written
There is no generated reference in the kit today — no OpenAPI document, no error catalog, no route table dumped at build time.
So where a page here disagrees with the source, the source is right, and the difference is worth reporting.
Two pages say so in their own words: error codes and HTTP routes. The kit does hold its routes to a contract in its own tests — declared routes checked against what actually mounted, in both directions — but that check does not reach these pages.
Dated claims
Anything about a price or a competitor carries an as-of date, because both change and neither is ours to control. Where you see one, re-check before you decide on it.