Skip to content

Guides

9 guides, each deep-linkable at the section level — so start at the top if you have not made a call yet, or jump straight to the question you arrived with.

Quickstart Three calls from an empty account to a computed ARR waterfall, on a test key. Start →

The guides

Each one is a single topic taken to the end. The questions under each are links into the exact section that answers them.

The three objects the whole API is built from, and the one idea — the customer-month fact — that explains most of its behaviour.

Choosing an input model, what each one forecloses permanently, and validation designed so a bad file costs you one round trip rather than a hundred.

The response contract in full. Read this one carefully — a client written against it correctly will not need rewriting when we add metrics.

The three credentials this API accepts, which of them you can get without asking us, and what each 401 is actually telling you.

One problem shape across every endpoint, what idempotency guarantees when you retry, and the single case where the API cannot tell you which of two things went wrong.

Exports 6 min

PDF, Excel and JSON artifacts: the asynchronous flow, the two different lifetimes on a finished export, and the four ways the endpoint refuses.

The exact billing formula, why it is a rectangle rather than a sum of contract terms, and why nothing is ever refused for being expensive.

The two key prefixes, what a test account actually starts with, and the boundary between the modes — which is the one place in this API where nothing is allowed to leak.

Connect this API to Claude, hand over contracts in a conversation, and get the same numbers your dashboard shows — with a report at the end of it.

What lives elsewhere

Two things are deliberately not in these pages, because keeping a second copy of either is how documentation starts lying.

  • The API reference — every endpoint, parameter, status and schema, rendered from the OpenAPI document. That document is still hand-authored: the service does not emit it yet, and nothing mechanically diffs the two. Where it and these guides disagree, the guides were written against the running service.
  • The metric knowledge base — what each metric means, how it is defined, and what a healthy value looks like. That is a finance question rather than an integration question. It lives on the main site, not here.

Not written yet

These are planned and named here so you know whether to wait or to ask.

One of them is worth answering now rather than later: there is no rate limiting — no 429, and nothing rejects a query for being over budget. The limits that do bind are on upload size: 10MB inline, 20MB by workbook, 50,000 rows.

  • Find winning segments — the combinatorial segment sweep — not yet built, so not yet documented