> ## Documentation Index
> Fetch the complete documentation index at: https://docs.sdk.anghami.com/llms.txt
> Use this file to discover all available pages before exploring further.

# llms.txt

> The LLM-friendly briefing file at /.well-known/llms.txt — what's in it, how to use it.

[`/.well-known/llms.txt`](/.well-known/llms.txt) is a single markdown document optimized for being dropped into an LLM context window. It follows the [llmstxt.org](https://llmstxt.org) convention.

## What's in it

* **Identity** — what the API is, who it's for, the base URL.
* **Authentication** — API key vs OAuth, header names, scopes.
* **Pagination** — the cursor scheme.
* **Service inventory** — one line per service with its purpose.
* **Pointers** — the bundled OpenAPI URL, the docs site URL, this page.

The goal is to give an LLM enough scaffolding in **a few hundred tokens** to plan correct calls without reading the full OpenAPI (which is tens of thousands of tokens).

## How to consume it

### As a system prompt component

```ts theme={null}
const llmsTxt = await fetch("https://docs.sdk.anghami.com/.well-known/llms.txt").then(r => r.text());

const messages = [
  {
    role: "system",
    content: `You are integrating with the Anghami + OSN+ SDK. Reference:\n\n${llmsTxt}`,
  },
  { role: "user", content: userQuery },
];
```

### For agent planning

Use `llms.txt` to plan. When a specific operation is needed, fetch the relevant slice of the OpenAPI bundle to get exact request/response shapes — don't load the full bundle into context.

### For documentation grounding

Mintlify's built-in "Ask AI" indexes both the MDX content of this site and the bundled OpenAPI — so on this docs site, you don't need to wire up `llms.txt` yourself. It's there for **other** agents that aren't already integrated with our docs.

## Caching

The file is served with `Cache-Control: public, max-age=3600`. Memoize for an hour; refetch on cache miss. We update it whenever the service inventory or auth shape changes.

## Why not just use OpenAPI?

OpenAPI is the **machine** contract — exact, exhaustive, large. `llms.txt` is the **LLM** briefing — concise, opinionated, readable.

An LLM that sees only the OpenAPI has to figure out the high-level shape of the API from the operations alone, which wastes tokens and produces hesitant tool calls. An LLM that sees `llms.txt` first knows what the API is *for*, and uses the OpenAPI to fill in the exact details.

Use both: `llms.txt` for planning, OpenAPI for execution.