Skip to main content
The whole public docs portal is also published as plain text for agent consumption, following the llms.txt convention.

What’s available

llms.txt — index

Hierarchical link index of every doc page. Compact (~30 KB). Best when you want the agent to navigate to specific topics.

llms-full.txt — flattened

Every page concatenated as a single document (markdown). Larger (~1-3 MB depending on size). Best for stuffing the agent’s context once and answering many questions.
Both are auto-regenerated by Mintlify on every docs deploy — they always reflect the current state of docs.getsly.ai. Each individual page is also available as raw markdown by appending .md to its URL:

How agents typically use this

Pattern 1 — knowledge base for support agents

Refresh once per deploy or on a daily cron.

Pattern 2 — targeted lookup via index

Cheaper on tokens than stuffing all docs upfront.

Pattern 3 — MCP server consumption

If you’re using the Sly MCP server, it exposes documentation lookup as a built-in tool. Agents call sly_docs_search(query) and get the relevant page back as a tool result.

Caching guidance

  • llms.txt changes when the doc structure changes (rare). Cache for a day.
  • llms-full.txt changes on every doc deploy. Cache for an hour.
  • Per-page .md changes when that page is edited. Cache for an hour.
Both files are served with appropriate Cache-Control headers. Sly is agent-first across multiple surfaces:
  • MCP catalog — every Sly capability as an MCP tool, drop-in for Claude Desktop / Cursor / Windsurf
  • A2A agent card — public, machine-readable identity + skill catalog
  • OpenAPI specdocs.getsly.ai/api-reference/openapi.json for codegen
  • Function-calling tool definitionsGET /v1/capabilities/function-calling returns OpenAI/Anthropic-compatible tool defs
Use whichever surface fits your runtime.