Skip to content
All servicesBuild

API development

An interface other systems can integrate with. Documented and versioned.

  • REST and GraphQL
  • Auth and rate limits
  • Documentation
  • Webhooks and events

What this is for

Data that is read and written by another system rather than by a person. An API is a contract: once something has been built against it, changing it breaks somebody else’s software rather than your own. That is why its design, its documentation and its versioning matter more here than the implementation behind them. It is what you build when a partner, a mobile app, or a second system of your own needs the same data as the first.

What it covers

  • REST and GraphQL

    Whichever suits what is consuming it — REST where the consumers are many and each does something simple, GraphQL where one client needs to ask for varied shapes of the same data.

  • Auth and rate limits

    Keys or tokens issued per consumer, with limits set so that one client cannot degrade the service for everybody else.

  • Documentation

    Written from the endpoints themselves so it cannot drift away from the behaviour, with examples that can be run rather than only read.

  • Webhooks and events

    For the cases where a consumer needs to be told that something happened, rather than repeatedly asking whether it has.

What you end up with

  • Documented endpoints

    Each one described with its parameters, its responses and its errors — including the failure cases, which is the half most documentation omits.

  • Keys, auth and rate limits

    Issued per consumer, so access can be revoked for one of them without affecting the rest.

  • A sandbox to build against

    A separate environment with representative data, so an integrator is not testing against live records.

  • Versioning, so an update breaks nothing

    New versions published alongside the existing one, with a stated path off the old.

What we need from you

  • Who is going to consume it

    An API designed without a known consumer is a guess. Even one real integrator changes the design for the better.

  • The data model as it currently is

    The API has to expose what exists. Tidying the model first is a legitimate decision, but it should be taken knowingly rather than discovered halfway.

  • What must never be exposed

    The fields that are internal, and the ones that are personal data, named before anything is published.

Who works on it

How it runs

  1. DiscoveryFirst

    We work out what you need, what it has to do, and what it will cost.

    • Scope written down
    • Fixed quote
    • Agreed timeline
  2. DesignBefore the build

    You see the site before it is built, on desktop and phone, and approve it before the build starts.

    • Designs to approve
    • Mobile and desktop
    • Final copy
  3. BuildThe build

    The design is built to its measurements, on a link you can open at any point during the build.

    • Live preview link
    • Progress you can see
    • Content loaded
  4. LaunchHandover

    Domain, hosting and email pointed at the new site. Checked on real devices, handed over in your name.

    • Live on your domain
    • Accounts in your name
    • Support after launch
  5. SupportAfter launch

    Updates, fixes and changes for as long as the site is live.

    • Updates and fixes
    • Monitoring
    • A number to message

Questions people ask first

  • With a conversation about what you actually need, and then a written scope with a price against it. Nothing is built until you have agreed both, and that scope is what the work is measured against for the rest of it.

Also under Build

Get in touch

Every one of these reaches us directly, and is answered by a person.

11:15Triple A