Machine-Readable Docs

Feed Across documentation directly to AI agents and RAG pipelines via llms.txt endpoints.

Every page on this site is available in machine-readable markdown format. AI agents, RAG pipelines, and LLM tools can consume Across documentation without scraping HTML.

Endpoints

Endpoint	Content	Best For
`/llms.txt`	Page index with titles, URLs, and descriptions	Directory lookup — find the right page to read
`/llms-full.txt`	All pages concatenated as markdown	RAG ingestion, large-context agents
`/docs/<path>.mdx`	Single page as raw markdown	Targeted reads — fetch exactly one page

When to Use Which

Agent needs to find a page — fetch /llms.txt, scan for the relevant title, then fetch that page's .mdx endpoint.
Agent needs full context — fetch /llms-full.txt and pass the entire corpus (or relevant chunks) into the context window.
Agent already knows the page — append .mdx to any docs URL to get raw markdown. For example, /docs/ai-agents/llms-txt.mdx returns this page as markdown.

The /llms-full.txt endpoint is large. If your agent has a limited context window, prefer /llms.txt to find the right page and then fetch it individually via .mdx.

Fetching Examples

Page Index

terminal

curl https://docs.across.to/llms.txt

agent.ts

const index = await fetch("https://docs.across.to/llms.txt").then(r => r.text());
console.log(index);

agent.py

import requests

index = requests.get("https://docs.across.to/llms.txt").text
print(index)

Full Documentation Dump

terminal

curl https://docs.across.to/llms-full.txt

agent.ts

const full = await fetch("https://docs.across.to/llms-full.txt").then(r => r.text());
console.log(full);

agent.py

import requests

full = requests.get("https://docs.across.to/llms-full.txt").text
print(full)

Single Page

terminal

# Fetch the Swap API docs as markdown
curl https://docs.across.to/docs/introduction/swap-api.mdx

agent.ts

const page = await fetch("https://docs.across.to/docs/introduction/swap-api.mdx")
  .then(r => r.text());
console.log(page);

agent.py

import requests

page = requests.get("https://docs.across.to/docs/introduction/swap-api.mdx").text
print(page)

Per-Page Actions

Every documentation page includes built-in buttons for AI consumption:

Copy Markdown — copies the page content as markdown to your clipboard
Open in Claude / ChatGPT / Cursor — opens the AI tool with the page URL pre-loaded as context

These buttons appear at the top of each page and use the same .mdx endpoints described above.

For RAG Pipelines

The /llms-full.txt endpoint concatenates all pages with # Title headings. You can split on H1 headings to chunk by page:

chunker.py

import requests

full_text = requests.get("https://docs.across.to/llms-full.txt").text

# Split into per-page chunks
chunks = []
current_chunk = ""

for line in full_text.split("\n"):
    if line.startswith("# ") and current_chunk:
        chunks.append(current_chunk.strip())
        current_chunk = line + "\n"
    else:
        current_chunk += line + "\n"

if current_chunk.strip():
    chunks.append(current_chunk.strip())

print(f"Split into {len(chunks)} page chunks")

Each chunk corresponds to one documentation page and can be embedded independently for retrieval.

Machine-Readable Docs

On this page