Building with AI

grandocs is built agent-first. There are four ways in — use whichever your tooling supports.

1. llms.txt

Three machine-readable indexes, generated at build time:

/llms.txt — curated index with usage instructions
/llms-full.txt — the full corpus (latest versions) inlined
/llms-small.txt — a compact variant for small context windows

The llms.txt preamble tells agents the conventions below: default to grandMA3 2.4 but honor the user’s actual console version, and always cite the canonical help.malighting.com page.

2. Markdown twins

Append .md to any docs URL to get the raw markdown — the same convention Stripe uses. For example:

https://grandocs.samcarlton.com/grandma3/2-4/cue_store/      → the page
https://grandocs.samcarlton.com/grandma3/2-4/cue_store.md    → raw markdown

Every page also carries a Copy as Markdown button and Open in ChatGPT / Claude links in its header.

3. MCP server

A stateless Model Context Protocol server over grandocs’ search index. Tools: search_docs, get_page, command_lookup, and list_products_versions.

# Claude Code
claude mcp add --transport http grandocs https://grandocs-mcp.samcarlton.com/mcp

Cursor, Claude Desktop, and other clients: see the MCP server README.

4. Skills

Installable agent skills for grandMA documentation work — for Claude Code, Cursor, and 20+ other agents:

npx skills add ThatGuySam/grandocs

Includes grandma3-docs, grandma2-dot2-docs, grandma-command-syntax, grandma3-lua-api, grandma-macro-writer, and grandma-troubleshooting.