TL;DR: md2wechat v2.1.0 ships a Layout system with 43 built-in modules, a six-dimension Authentic Humanize engine, and native Agent support. It's no longer just a Markdown-to-WeChat formatter — it's AI content production infrastructure.

---

§The Short Version

A month ago md2wechat crossed 1.6k stars. I'd been thinking about where to take it next: keep it as a formatting tool, or push further?

v2.1.0 answers that question. Writing, humanization, layout design, and publishing — one pipeline, one pass.

---

§Layout System: AI-Driven Typesetting

This is the biggest update.

Before v2.1.0, you wrote Markdown, picked a theme, and md2wechat converted it. Layout was still on you.

Now there's a Layout system with 43 built-in modules across 7 categories: attention-grabbing openings, opinion blocks, data infographics, brand expression, conversion sections — covering the common content scenarios you find in WeChat Official Account articles.

Usage is straightforward:

md2wechat layout list      # list available modules
md2wechat layout show <id> # preview a module
md2wechat layout render    # render

There's also a validate command that checks formatting before rendering.

AI doesn't just write articles anymore. It handles typesetting design too.

---

§Authentic Humanize: Six Dimensions of De-AI

Most de-AI tools swap高频 words or rearrange sentence structures. The effect is limited.

md2wechat's Humanize works across six dimensions: vocabulary, sentence patterns, tone, expression style, structure, and overall texture. It doesn't just replace a few words — it makes content read like a professional human writer wrote it.

md2wechat humanize --intensity authentic

Three intensity levels: gentle, medium, authentic. Authentic is the strongest and the closest to genuine human output.

In my own testing, content processed in authentic mode reads noticeably more natural. Not "AI text that's been tweaked" — actually like a human wrote it.

---

§Native Agent Support

This one is for developers.

md2wechat now supports JSON-first I/O, includes validate and render interfaces, and modules can be auto-discovered. You can call md2wechat directly from Claude, Codex, Cursor, GPT Agent, or any AI workflow tool.

The old content pipeline:

Write → Format → Adjust → Publish. Four steps, lots of context switching.

Now: AI completes writing + humanization + layout + publishing in one pass.

---

§Who Is This For

Content creators: Save time on formatting and polishing.

WeChat team operators: Humanize and Layout keep content quality consistent.

AI workflow developers: Native Agent support lets you treat md2wechat as a node in your content production pipeline.

Independent developers: PRs welcome. MIT licensed.

---

§Final Thoughts

After building content tools for this long, one thing keeps getting clearer:

The end goal of a tool isn't feature accumulation — it's freeing creators to spend time on what actually matters.

Formatting shouldn't take an hour. Polishing shouldn't take an hour. Those things belong to the tool.

Creators should spend their time on topic selection, viewpoints, and insights. That's the irreplaceable part.

v2.1.0 is live on GitHub.

GitHub: github.com/geekjourneyx/md2wechat-skill