LLM Wiki — Synaptic Lattice

docmd.io

Properties1
tagsentity, tool, llm-wiki, anthropic, standard

Leitura de 3 min

Overview (confidence: high)

docmd.io is an open-source (MIT) documentation engine that converts Markdown files into production-ready documentation sites. It is relevant to this vault because it ships native AI-agent support: a built-in MCP server, auto-generated llms.txt / llms-full.txt context files, and a SKILL.md instruction file created by docmd init for coding agents. It was surfaced in the Raindrop “Second Brain” collection as a documentation tool for AI-powered workflows.

Key Facts (confidence: high)

  • License / model: Open source, MIT, zero-config. Produces standalone static HTML with <20kb client JS; Lighthouse 100.
  • AI-agent support: ships a native MCP server (docmd mcp) so agents can search, read, and validate docs live; auto-generates llms.txt context; docmd init scaffolds a SKILL.md instruction file for coding agents (Claude, ChatGPT, Gemini, DeepSeek, Agent Skills).
  • Content: rich Markdown containers (callouts, tabs, cards) with no HTML; native i18n, light/dark themes, multi-version docs, OpenAPI support.
  • Deploy: one-command to GitHub Pages, Docker, Nginx, or Caddy.
  • Repo: https://github.com/docmd-io/docmd

Features

Core Architecture

  • Zero-config static site generator — produces standalone HTML from markdown
  • Lighthouse 100 — perfect performance score
  • <20kb client JS — minimal frontend footprint
  • Multi-version docs — version management built in

Ingest & Processing

  • Markdown-native — reads .md files directly
  • Rich containers — callouts, tabs, cards without HTML
  • OpenAPI support — auto-generates API docs from specs

MCP / Tools

  • Native MCP server — docmd mcp exposes docs to agents
  • llms.txt + llms-full.txt — auto-generated agent context files
  • SKILL.md scaffolding — docmd init creates agent instruction files
  • Live validation — agents can validate docs in real-time

Quality & Governance

  • Lighthouse 100 — perfect performance
  • i18n — multi-language documentation
  • Light/dark themes — built-in theme support

Integration

  • Agent Skills — works with Claude, ChatGPT, Gemini, DeepSeek
  • GitHub Pages — one-command deploy
  • Docker/Nginx/Caddy — multiple deploy targets
  • MCP protocol — standardized agent access

Data & Storage

  • Static HTML — no server runtime required
  • Markdown source — plain text authoring
  • Standalone output — self-contained HTML files

Relationships (confidence: high)

  • “operates-with” model-context-protocol — docmd exposes docs to agents via a native MCP server.
  • “compatible-with” claude-skills — docmd init generates a SKILL.md instruction folder for coding agents.
  • “sibling-of” open-knowledge-format — llms.txt is a sibling “agent context” standard; docmd is a publish path for Markdown knowledge.
  • “operates-with” llm-wiki — a wiki is itself a Markdown corpus that docmd-style tooling can expose to agents.
  • “sibling-of” agent-memory-systems — docmd is adjacent (artifact/docs exposure) rather than memory, but shares the agent-access theme.

Implications (confidence: high)

docmd shows a low-friction path to expose a Markdown corpus (including a wiki like this vault) to agents: MCP server + llms.txt means an agent could query and write our docs directly. It previews how the second brain’s content could be made agent-accessible without leaving Markdown.

Sources

^[raw/articles/raindrop-second-brain-collection-2026.md]

Visão de gráfico

Backlinks