For LLMs & agents

Every Hep.gg API doc ships a plain-text, machine-readable twin for LLMs, coding agents, and RAG indexers. Open it from the "LLM doc" button on any API page, or fetch it directly.

For LLMs & agents#

Every API reference in these docs ships an LLM twin: a second copy of the same page, rewritten for machines. The twin has the same facts as the human page but is stripped of styling and structured for parsing: flat headings, short declarative sentences, every field spelled out with its type and constraints, full request examples, complete response shapes, and every error code. Point a model, a coding agent, or a retrieval pipeline at a twin and it can integrate a Hep.gg app without scraping the rendered site.

Start at the index#

https://docs.hep.gg/llms.txt is the machine index: a plain-text glossary of every documentation page, each linked to its /llms/... endpoint. An agent can fetch this one URL and then recursively reach every page without ever leaving the /llms endpoints. Want it all in a single fetch instead? https://docs.hep.gg/llms-full.txt is the entire documentation set concatenated into one file.

Three ways to reach a page#

1. The "LLM doc" button. Every API reference page has one at the bottom. It opens that page's plain-text twin in a new tab. This is the easiest way to grab the reference for whatever you are reading.
2. Fetch it directly. Swap /docs for /llms in any page URL. The human page https://docs.hep.gg/docs/database/overview has the plain-text version at https://docs.hep.gg/llms/database/overview. This works for every page, not just the per-app overviews.
3. Start at the index and follow the links, as above.

# The index of every page
curl https://docs.hep.gg/llms.txt
 
# One page, as plain text
curl https://docs.hep.gg/llms/database/overview
 
# The entire documentation set in one file
curl https://docs.hep.gg/llms-full.txt

const index = await (await fetch("https://docs.hep.gg/llms.txt")).text();
const dbDoc = await (await fetch("https://docs.hep.gg/llms/database/overview")).text();
// Feed either straight into your model's context.

Legacy URLs still work#

Hep.gg used to publish these references at https://hep.gg/llms.txt and https://hep.gg/llms/<app>.txt. Those URLs keep working: each one now permanently redirects to its new home. https://hep.gg/llms.txt forwards to the index at https://docs.hep.gg/llms.txt, and https://hep.gg/llms/database.txt forwards to https://docs.hep.gg/llms/database/overview. Nothing you have bookmarked or cached breaks. Prefer the new URLs for anything new.

The directory#

The full index, listing every twin and the shared conventions, is itself a twin: llms/overview.

Conventions#

Every reference follows the same rules, so a model only has to learn them once:

• Base URL is https://hep.gg unless the file names a per-app host (for example sms.hep.gg, music.jakeypri.me, paste.hep.gg, forms.hep.gg).
• Auth is per-app and stated at the top of each file: a bearer key, an API key, a username plus token, or OAuth tokens. Cookie-authenticated endpoints exist for the dashboard and are listed only for completeness, marked as not for programmatic use.
• Responses are JSON envelopes: { "ok": true, "data": ... } on success and { "ok": false, "error": "<message>", "code"?: "<token>" } on failure. A few legacy endpoints, and the OpenAI- and Anthropic-shaped AI endpoints, sit outside the envelope and say so.
• Timestamps are ISO 8601 UTC.
• Rate limits, where present, are documented per endpoint and return HTTP 429 with Retry-After.

The twins are cached for a few minutes and are safe for crawlers and RAG indexers to fetch on a schedule.