跪拜 Guibai
← All articles
Frontend · Backend · GitHub

Archify Hits GitHub Trending as a Mermaid Alternative That Generates Cleaner Diagrams from Plain English

By 也无风雨也雾晴 ·
Read original on juejin.cn ↗ Google Translate ↗ Alt translation

Archify fills a real gap: Mermaid's output often looks utilitarian, and teams that need polished, shareable architecture diagrams without leaving the editor now have a zero-dependency alternative that runs as an agent skill inside the tools they already use.

Summary

Archify is an Agent Skill for Claude, Codex, and opencode that generates system and flow diagrams from plain English. It produces a JSON intermediate representation, runs schema and layout checks, and outputs a self-contained HTML file with dark/light themes and export to PNG, JPEG, WebP, and SVG. The project started in April with only architecture diagrams and grew to support four chart types, 4x resolution export, a CLI, CI, and a zero-dependency install package.

Organic sharing on Twitter drove its spread, with users noting the output looks better than Mermaid and the HTML format makes diagrams easy to distribute. Trendshift records show it first entered GitHub Trending on June 30, peaking at number 9.

The creator, a self-taught front-end engineer with a vocational-college background, built the tool while working at a startup and later at MiroFish. He credits the project's visibility with opening doors that a top-up bachelor's degree had previously closed, including passing an interview at a large Shanghai firm after failing two coding questions but demonstrating his open-source work.

Takeaways
Archify generates architecture, sequence, flowchart, and ER diagrams from natural language, outputting a standalone HTML file.
Exports include PNG, JPEG, WebP, and SVG with 4x resolution and dual light/dark themes.
A JSON Schema and layout checker validate the intermediate representation before rendering, catching structural errors early.
The project ships as a zero-dependency npm package with a CLI and CI integration.
Organic Twitter sharing, not a launch campaign, drove adoption; the hero image and visibly better output than Mermaid were the main hooks.
Trendshift data confirms the repo first hit GitHub Trending on June 30, 2025, ranking 9th.
Conclusions

Archify's growth pattern — zero marketing, pure Twitter word-of-mouth — suggests developer tooling still spreads fastest when the output is immediately visually superior to the incumbent, not just functionally equivalent.

The creator's story highlights a persistent hiring filter in Chinese tech: top-up bachelor's degrees are often treated as a hard disqualifier at large firms, regardless of demonstrated ability, making open-source portfolio a practical workaround.

Building a tool as an Agent Skill rather than a standalone app lowers the adoption barrier; users stay inside Claude Code or Codex and treat diagram generation as just another agent capability.

The project's evolution from a single diagram type to a full multi-format exporter with CI and a CLI in under three months shows how quickly a solo developer with AI-assisted coding can iterate when the scope is tightly defined.

Concepts & terms
Agent Skill
A capability extension for AI coding agents like Claude Code or Codex that lets the agent perform a specific task — in Archify's case, generating diagrams from natural language — without leaving the agent's workflow.
JSON IR (Intermediate Representation)
A structured JSON format that sits between the user's natural-language input and the final diagram output, allowing schema validation and layout checks before rendering.
Top-up Bachelor's Degree (专升本)
A Chinese education pathway where vocational college graduates complete an additional two years at a university to earn a bachelor's degree; often stigmatized by employers compared to a traditional four-year degree.
Source: juejin.cn ↗ Google Translate ↗ Backup ↗