FlowFinds

    Press Esc to close · ⌘K opens search anywhere

    Log inTry FlowFinds ↗
    Menu

    Documentation

    Documentation

    Build on FlowFinds: quickstart, concepts, API reference and guides.

    FlowFinds is an HTTP API over a running commerce operation. It is a JSON interface, served from https://flowfinds.ai, and it is the same interface the FlowFinds console itself calls — there is no privileged internal API behind it.

    These pages document what the server does, not what it might one day do. Where a mechanism does not exist yet, the page says so rather than describing a plausible version of it. Two examples you should read before you plan an integration: there are no API keys (see Authentication), and there is no versioned URL prefix (see Versioning and deprecation).

    Three ways in

    Quickstart

    You want to see it work. Sign in with a link, then read your own dashboard over HTTP. Fifteen minutes, one email address, no key to provision.

    Concepts

    You want to know what you are modelling. Product, offer, store, claim, intent, evidence, verdict, referral — the eight nouns every endpoint is made of, and how they constrain each other.

    API reference

    You are already building. Every route, its method, what it takes, what it returns, and which refusal it can answer with.

    What the API is shaped like

    Four properties are worth knowing before you read any endpoint page.

    • The session is a cookie. Every response can set ff_session; every request should send it back. There is no bearer token and no key to rotate.
    • A refusal is usually a 200. Many endpoints answer a request they will not perform with {"status":"refused","reason":…} at HTTP 200, because the request was well-formed and the answer is a fact about your account. Branch on reason, not only on status.
    • The refusal vocabulary is enumerable. GET /api/reasons returns every reason code the server can emit, with its explanation. Your error handling can be generated from it.
    • Nothing under /api/ is cacheable. Those responses carry Cache-Control: private, no-store, must-revalidate, because they are per-account state and an edge cache would serve them to the wrong account.

    Every page in this section

    DevelopersThe FlowFinds API: one HTTP interface, the same one the console uses, documented endpoint by endpoint.QuickstartFrom an API key to a first successful call.ConceptsProducts, offers, stores, claims, intents and evidence — the model behind the API.AuthenticationAPI keys, scopes and how requests are authorised.Rate limits and quotasPer-endpoint limits, headers, and what happens when you exceed them.ErrorsEvery error code the API returns, and how to handle each one.API referenceEvery endpoint, parameter, field and response.Agents and toolsThe tools the agent can call, and how to extend them.GuidesTask-shaped walkthroughs for the things people actually build.CookbookWorking recipes you can paste and run.SDKs and CLIOfficial libraries and the command line.TroubleshootingSymptoms, causes and fixes for the failures people hit.GlossaryEvery term FlowFinds uses, defined once.Versioning and deprecationHow the API changes, and the notice you are guaranteed.ChangelogDated record of every change to the API and the product.Migration guidesMoving between API versions without breaking.

    Outside these pages

    The API is one face of the product. For what the agent does and why, read the product pages. For how well it does it, measured rather than asserted, read the research. For the commitments that sit around the API — retention, sub-processors, uptime — read Trust.