# How agents access 2plot documentation

> The network's standing policy for AI agents and crawlers — what is always free, what a free account unlocks, and how the metered lane will work when it arrives.

**Site index:** [https://2plot.ai/llms.txt](https://2plot.ai/llms.txt) — every page on this site, as Markdown.  
**Sibling sites:** 16 more in The 2plot network — listed in the site index above.  
**Sitemap:** https://2plot.ai/sitemap.xml  


---


## The short version

Every app in the 2plot network publishes machine-readable documentation —
`llms.txt` on every host, a markdown twin for every page. This page is the
policy for reading it. Three lanes:

1. **Free, always** — discovery and adoption. `robots.txt`,
   `sitemap.xml`, every root `/llms.txt` index, install instructions,
   and each package's compact briefing. Enough to find a component,
   install it, and start building. This floor is permanent and loud:
   the documents exist to get the packages used.
2. **Free with an account** — everything. A free 2plot.ai account gives
   you (and your agent) an agent key that travels in the URL and unlocks
   the full documentation corpus on every satellite: exhaustive prop
   tables, callback patterns, full-depth prose. Sign in once at
   [2plot.ai](https://2plot.ai); the key follows the links.
3. **Metered, coming** — anonymous bulk. Crawlers that pull entire
   corpora at scale without identifying themselves will meet an HTTP 402
   with a machine-payable day pass (per package, or one bundle for the
   whole network). Not live yet — the price arrives after the policy,
   not before.

The principle behind lane 3: the paid path is for callers that **choose
not to identify**. Identify — as a person or as an agent holding a key —
and the network is free.

## Sizes: pick the document that fits your context

Rolling out across every host in the network (the Svelte convention):

| Document | What it is |
|---|---|
| `/llms-small.txt` | A compact briefing — what the package is, install line, one line per page. Start here if context is tight. |
| `/llms.txt` | The index — tagline, page directory with per-page markdown twins, network links. |
| `/llms-full.txt` | Every page's prose in one document, source-annotated. The whole manual in one fetch. |

Every documentation page also serves its own twin at
`/<page>/llms.txt` — fetch exactly what you need.

**Prefer one full-corpus fetch only when the task needs the whole manual
and the document fits your context.** Sizes differ by an order of
magnitude across the network: on the hub `/llms-small.txt` is about 3 KB,
`/llms.txt` 11 KB and `/llms-full.txt` 22 KB, while a component satellite
runs 4–6 KB, 12–43 KB and 100–520 KB respectively (Leaflet's full manual
is 521 KB, Flows' 426 KB, Pannellum's 98 KB — measured 2026-09-08). Take
the smallest sufficient document: the compact briefing to choose a
package, the index to find the page, a single page's twin for one
component's props, and the full corpus only for work that spans the whole
package. Every compact briefing carries the package's install line, so an
agent that reads only the smallest document can still install the right
distribution.

## Why this policy

These docs are the top of an open-source funnel: an agent that reads
them and helps a developer ship with `dash-leaflet2` or `dash-email` is
the system working as intended. What the metered lane prices is not
knowledge but bulk — the training-crawler pattern of thousands of
fetches that never refer a single visitor back. Humans and their agents
ride free; volume without identity pays for the bandwidth it uses.

Questions, or want your crawler allowlisted? The network's front door is
[2plot.ai](https://2plot.ai) — every satellite links back here.


---

*Source: /docs/agent-access*