Markdown and llms.txt

Every Fillo docs page is also published as machine-readable Markdown, so a coding agent can pull stable product context without scraping page chrome.

Per-page Markdown

Append .md to a docs or guide URL:

https://fillo.so/docs/embed.md
https://fillo.so/docs/fields/file-upload.md
https://fillo.so/guides/nextjs-form.md

Each response is served as text/markdown, points back to the canonical web page, and is marked noindex, follow so the HTML page stays the search result. New MDX pages include their validated frontmatter and article source; legacy pages use their existing structured Markdown builders.

Agent indexes

Start with /llms.txt and fetch only the per-page Markdown you need. Reach for /llms-full.txt when a tool works better with one complete context file — and its context window can actually hold it.

Example retrieval

curl -L https://fillo.so/llms.txt
curl -L https://fillo.so/docs/response-limits.md
curl -L https://fillo.so/openapi.json

const docs = await fetch("https://fillo.so/docs/webhooks.md").then((r) => {
  if (!r.ok) throw new Error(`Docs request failed: ${r.status}`);
  return r.text();
});

Keep agent work grounded

Ask the agent to cite the docs route and check the installed package version before it changes your application. Public docs describe the current Fillo service; your project may pin an older SDK that doesn't have a newer prop or field behavior yet.

Reading the docs doesn't grant an agent any access. It still needs the right publishable key, login token, or scoped workspace key, and the server applies the same validation and workspace boundaries either way.

Related

• MCP server: Let an agent retrieve docs and operate through typed tools.
• SDK reference: Look up the installed renderer and controller contracts.
• Security and boundaries: Keep management credentials and respondent data in the right environment.