The Inspired library.

The Inspired integration library

Everything published for building on the Inspired platform, in the order it is worth reading. Six works, six formats, one namespace.

This is a collection: it has no content of its own beyond this page and the reading order below. Each member is an independently published resource with its own page, its own versions and its own downloads.

The order

| # | Work | Format | Read it when | | --- | --- | --- | --- | | 1 | Shopping a photograph | guide | You have a key and have to ship a feature | | 2 | The Inspired Handbook | book | You want the reasoning behind the design | | 3 | inspired CLI | cli | You want a working read from a terminal today | | 4 | Inspired visual search skill | skill | An agent will be handed photographs | | 5 | Ask for the shirt, get the vest | article | You are building your own extraction pipeline | | 6 | Focus-domain dataset | dataset | You are mapping the response into your own schema |

Nothing here is required in order to call the API — the platform's own quickstart is enough for a first read. Start here when it has to work unattended.

Two routes through it

Shipping an integration. The guide (1), then the handbook's chapters 3, 5 and 7 (2) while you write the client, then the dataset (6) when you map the response. The CLI (3) is worth running first regardless: one command gives you a real payload to develop against, which saves spending trial requests on debugging.

Running an agent. The skill (4) plus the handbook's chapter 6 (2), which is what the MCP transport does and does not implement. The guide (1) is still worth an hour — an agent driven by a developer who has read it asks better questions.

Neither. If you are here because you extract objects from photographs and want to know how somebody else's prompts are shaped, read the article (5). It stands alone and needs no key.

What this collection is not

It does not bundle anything. Listing works as members creates a reading order and a set of links, not a download — the collection page will not hand you an archive of all six.

The downloadable kit is a separate thing, declared by the Inspired product and by the tool: a versioned ZIP pinning the skill, the CLI, the guide and the dataset at specific versions. If that is what you want, it is on either of those two pages, not this one.

Versioning

Members are pinned by identity, not by version: this page always points at the current release of each work. The handbook and the guide move together when the platform changes; the article is a field report and is only revised if its conclusions change.

I'm picking up an Inspired integration next week and I have about three hours
before the kickoff. Here is the Inspired integration library — six works.

Plan my three hours, and then execute the reading for me:

1. Look at the collection's reading order and its two routes. I am shipping a
   server-side integration, not running an agent, so tell me which members I can
   skip entirely and say why rather than hedging.
2. Read what you recommend and give me a one-page briefing: how the pipeline
   works, why a read is minutes, how the allowance is spent, and what the result
   object contains.
3. Pull out every constraint that would change an architectural decision — the
   things I would be annoyed to discover in week two. I am expecting something
   about durable jobs and something about when quota is charged.
4. Give me the ten-point checklist out of the guide as a checklist I can paste
   into our tracker, with one line each on what it means.
5. Tell me the vocabulary rules from the handbook's last chapter as a list of
   banned and permitted words, because I will have to argue about that with
   whoever writes our UI copy.

Skip the article unless something in it changes how I should call the API — I
am not building an extraction pipeline. And do not summarise the dataset's
columns at me; just tell me whether I will need it and when.