by che1sy
Auto-convert pasted LLM output (ChatGPT / Claude / Gemini) into Obsidian-flavored Markdown: math delimiters, headings, citations, tables.
mdfit - Fit Any Markdown
An Obsidian plugin that auto-converts pasted LLM output (ChatGPT / Claude / Gemini) into clean Markdown the moment it lands in a note — no manual cleanup.
Paste from an AI chat and get, instantly:
- Math that renders:
\(...\)→$...$,\[...\]→$$...$$, LaTeX environments wrapped, mid-sentence display math kept on its own lines. - Correct heading levels (ChatGPT's shallow
###headings promoted to#). - Citation noise like
【20†L33-L40】and dangling[1]stripped. - Real Markdown tables and task lists recovered from chat-app HTML.
**Note:**paragraphs turned into Obsidian callouts.- Tidy CJK ↔ Latin spacing, zero-width characters and "Copy code" junk removed.
Code blocks, inline code and math are never touched — conversion is guarded by a protection layer, so examples inside your notes stay byte-identical.
Usage
Just paste into any note (Ctrl+V). The plugin converts the clipboard content
(plain text, or the rich HTML flavor when it carries tables/KaTeX) and inserts
the result. Toggle it at any time with the command
"mdfit: Toggle paste conversion on/off".
Install
From the community directory (recommended)
Settings → Community plugins → Browse → search "mdfit" → Install.
Manual install
- Download
main.jsandmanifest.jsonfrom the latest release. - Copy both files into
<your-vault>/.obsidian/plugins/mdfit/. - Enable mdfit - Fit Any Markdown in Settings → Community plugins.
Development
The source code lives in the
md-fit-all monorepo
(packages/obsidian-plugin), alongside the mdfit CLI and the conversion
engine mdfit-core. This repository is the release channel for the Obsidian
community directory.
MIT License.
For plugin developers
Search results and similarity scores are powered by semantic analysis of your plugin's README. If your plugin isn't appearing for searches you'd expect, try updating your README to clearly describe your plugin's purpose, features, and use cases.