Convert raw HTML to clean Markdown (with heading/link/code-block style options) via @mukundakatta/html-to-markdown-mcp (npx)

@mukundakatta/html-to-markdown-mcp v latest

Install: npm install @mukundakatta/html-to-markdown-mcp Entry: src/index.js (stdio) Tools: 2 — html_to_md, extract_text

Tool: html_to_md

Converts HTML string to Markdown.

Parameters:

• html (string, required) — the HTML input
• heading_style (enum: atx|setext, default: atx)
• code_block_style (enum: fenced|indented, default: fenced)
• bullet_list_marker (enum: -|*|+, default: -)
• link_style (enum: inlined|referenced, default: inlined)
• strip_scripts (boolean, default: true)

Returns: {"markdown": "..."}

Tool: extract_text

Strips ALL HTML tags, returns whitespace-normalized plain text.

Parameters: html (string, required) Returns: {"text": "..."}

Key gotchas

1. Tables render as flat text — no pipe-delimited markdown table syntax. <table> output is just cell text separated by double-newlines (Name\n\nAge\n\nAlice\n\n30). Major limitation for tabular data.
2. `strip_scripts: true` is default — <script> and <style> tags are removed. Set false to keep script text (rendered as plain text, not code blocks).
3. Setext headings produce underline-style (===== / ---) instead of # marks.
4. Referenced links produce numbered references at the bottom ([1]: https://...).
5. HR renders as * * * (not --- or ***).
6. First call ~167ms (JIT), subsequent calls 2-23ms. p50 = 9ms.
7. extract_text also strips script/style content — not just tags.
8. Fenced code blocks are triple-backtick style by default.
9. Images preserved as ![alt](src) format.
10. Blockquotes correctly render with > prefix.

Tested: 16 calls, 100% success, p50=9ms