LLM Text

Use llms.txt when your IDE, chat client, or coding agent can read web pages but cannot connect to the sdocs MCP server.

The site exposes three LLM-readable entry points:

• /llms.txt: curated index that follows the llms.txt proposal
• /llms.mdx/docs/...: markdown mirrors for individual guide and API pages
• /llms-full.txt: complete guide and API corpus for large-context agents

Recommended setup

Add the public sdocs URL as documentation context in your agent. Replace <docs-origin> with the origin where this documentation site is deployed:

<docs-origin>/llms.txt

If you are running the docs locally, use your local site origin instead:

http://localhost:4000/llms.txt

Then ask your agent to read /llms.txt first and fetch linked markdown pages as needed. This keeps context usage lower than pasting the full documentation corpus into every conversation.

Large context setup

Use /llms-full.txt only when your tool supports large context windows and can handle a full generated API reference in one request:

<docs-origin>/llms-full.txt

This is useful for one-shot documentation ingestion, but most coding tasks should start from /llms.txt and pull only relevant API pages.

Prompt template

Use this as a system or project instruction for non-MCP agents:

For s&box documentation, read <docs-origin>/llms.txt first.
Use the linked /llms.mdx/docs/... markdown pages as primary sources.
Prefer targeted API type pages over /llms-full.txt unless broad context is required.
When citing API details, include the type or member name from the markdown page.

Choosing between MCP and llms.txt

MCP is still the best option when your tool supports it because it can search, resolve symbols, inspect members, and return structured API metadata.

Use llms.txt when your tool only supports URL-based documentation context, custom instructions, or manual web fetches.

The format follows the llms.txt proposal: a markdown file at /llms.txt with a short project summary and curated file lists for agents to fetch at inference time.