Honest preview
A render that matches what the tokenizer parses — not a prettied-up illusion. Pseudo-tags and raw HTML are shown plainly.
Markdown for machine readers
Parchmint shows the honest preview an LLM actually sees — every token, every warning — so your prompts, skill files, and docs land exactly as written. No formatting surprises at inference time.
Free & open source · Local-first · No account, no telemetry
---
name: honest-preview
description: what you see is what
the model gets
---
## What the model reads
- [x] every token counted
- [ ] nothing hidden
Write **once**, ship clean.
Write once, ship clean.
01 — The idea
Most editors render Markdown for your eyes — pretty headings, hidden whitespace, silent assumptions. But your reader is a language model. Parchmint renders the document the way a tokenizer parses it, surfaces the token budget, and flags the structure that quietly breaks prompts. You stop guessing what the model sees. You just look.
## Context
You are a **precise** editor.
- [ ] verify each claim
02 — What's inside
A render that matches what the tokenizer parses — not a prettied-up illusion. Pseudo-tags and raw HTML are shown plainly.
Live token count as you type, with token-level warnings before a document blows past the budget you're writing for.
Document-type detection plus structural linting catches the broken outlines and malformed frontmatter that quietly derail prompts.
A live outline of every heading. Jump the cursor anywhere — or right-click the preview and edit the exact line you clicked.
Diagrams, math, and syntax-highlighted fences render inline. Three view modes — source, split, preview — a keystroke apart.
Built on Tauri and Rust. Your files stay on disk, external changes are detected, and nothing phones home. Ever.
03 — Take it home
Free, open source, MIT-licensed. Pick your platform — every build comes from the latest release.
Or build from source: npm install && npm run tauri build