llms.txt

The whole WOS API compressed into one plain-text file. Drop it into an IDE or a coding agent and it can build against WOS without reading anything else. Open the file.

How to use it

Point your tool at the URL, or paste the file into a project. It is plain text with no markup to strip, ordered so the first screen is the part you need first.

# fetch it
curl -s https://wontopos.com/llms.txt -o wontopos-llms.txt

# or hand the URL to an agent
"Read https://wontopos.com/llms.txt and add long-term memory to this app."

Every page on this site also points at it from the document head, so a crawler that reads the HTML finds it without being told.

What is in it

Fifteen sections, in this order. The 28 endpoints each get a line with their request and response shape, which is the part most tools need.

Start here

The two things that authenticate a call: an API key, and the store a call is scoped to.

Quickstart

curl and Python that run as written. Create a store, write a memory, recall it.

Endpoints

Every path with its request shape and response shape, one line each.

Engrams

Built-in multi-hop retrieval pipelines and when to reach for each.

Images

Storing a picture as a memory, the size window, what comes back, and how it is billed.

verify

Asking the engine again when one pass did not carry the answer.

Won

The free calls meant for a model to make on its own rather than for your UI.

Caching · Filters · Idempotency

Three optional switches: cheaper repeats, narrowing a search, and making your retry safe.

Speakers

Recording who said each memory, then recalling one person.

SDKs

Python, TypeScript and Rust — the same surface, released in lockstep.

MCP

Plugging WOS into tools you did not build, with per-client setup.

Common patterns

Which call to reach for in the situations that come up most.

Models & pricing

Every live model with its price, context size and benchmark score.

Errors & limits

What each status code means and what to do about it.

Trust

What we do and do not do with stored data, and the sub-processors involved.

How it stays current

The file is the source of truth for the API surface, and it carries the date it was last changed in its own header. The on-site AI guide can lag it by a day or two; the file does not. When something in the API changes, this file changes with the release rather than after it.

llms.txt or the docs?

• Give a machine llms.txt. One fetch, no navigation, no HTML. It is built for a tool that wants the whole surface at once.
• Give a person the docs. One page per endpoint, with the fields laid out, what comes back, and a curl you can paste.
• Give a program openapi.json. OpenAPI 3.1, for code generation and contract checks.

All three describe the same API. If they ever disagree, the OpenAPI spec is what the server actually enforces.