PrettyMD vs VS Code
VS Code ships the most widely installed Markdown preview in the developer world. I use it every week, and I still don't review agent-written specs in it. Here's where the built-in preview stops and where a dedicated read-only viewer takes over.
Why compare a free editor to a $19 reader
Because the two answers people give to “do I need a Markdown viewer?” are both half right. One camp says: open VS Code, hit Ctrl+Shift+V, done. The other says: your agent wrote 900 lines of spec overnight and reading it inside a code editor is where the review discipline goes to die. Both camps are describing different jobs, and almost nothing written about this comparison separates them. Consider this the separation.
My position, stated upfront so you can disagree with it early: VS Code is the finest free way to render Markdown, and it is the wrong place to review what an AI agent wrote. Not because the preview is weak. Because review is a distinct activity with its own requirements (memory of what you last read, feedback that survives without a chat window, a surface with no write tooling competing for attention), and a pane inside an editor meets none of them by design.
The glance table, before the argument
| PrettyMD | VS Code | |
|---|---|---|
| Best for | Reviewing .md files your agents produce | Writing and editing Markdown alongside code |
| Price | $19 once, 7-day trial, never a subscription | Free, open source (verified September 2026) |
| Standout capability | Line comments a compatible agent reads and acts on | A mature preview backed by the largest extension marketplace |
| Main trade-off | Read-only by design; Mac and Windows only | The render is a pane inside a full IDE; nothing tracks your reading progress |
| Markdown dialect | Mermaid, code blocks, full Markdown spec | CommonMark via markdown-it; Microsoft’s own FAQ says it does not target GitHub Flavored Markdown |
| Platforms | Mac and Windows | Mac, Windows, and Linux |
Where the built-in preview earns its reputation
Start with credit, because it’s overdue in most takes on this matchup. The preview renders instantly on any .md file with zero setup. Since v1.121, Mermaid fenced blocks draw real diagrams natively, no extension hunt. Point it at a git diff and it shows you the rendered document with changed lines highlighted, which is a genuinely elegant trick. And the ecosystem around it (linters, formatters, export pipelines) is unmatched at any price, let alone free.
If your Markdown week is mostly writing (READMEs you author, docs you lint and commit yourself), none of what follows will change your mind, and it shouldn’t. Stay in VS Code. It is the best tool for that job and it costs nothing.
The job the preview was never shaped for
Agent-authored documents break the preview’s model in three specific ways.
First, no memory. The preview renders what’s on disk right now. When Claude Code or Codex revises a spec you already read, the pane shows the new text with no signal about what moved. Git diff helps only if the agent’s work is committed, and in a typical vibe-coding loop most of it isn’t yet. You re-read 500 lines to find the 40 that changed.
Second, no feedback channel. You spot a wrong assumption in step 3 of a migration plan. In VS Code, that objection lives nowhere: you switch to the chat window and re-type which file, which section, what’s wrong. PrettyMD puts the comment in the document itself: highlight the line, write the note, and it lands in a readable yourfile.md.comments.json sidecar beside the Markdown. The viewer never touches the source; a compatible agent reads the thread, replies, and revises the Markdown itself. Feedback stays attached to the work.
Third, the wrong surface. A review session wants a reading tool: outline, tabs across the folder’s .md files, search, a mode that gets out of the way. The preview gives you a second pane in an editor built for typing code, with the activity bar, panels, and muscle memory that come with it. Fine for glancing. Costly for an hour of careful reading every morning.
Capability matrix: judged on review work
| Capability | PrettyMD | VS Code |
|---|---|---|
| Feedback that reaches the agent | PMD Line comments in a JSON sidecar the agent reads and answers | Nothing in the preview; objections travel through your chat window |
| Changes since your last read | PMD Per-file highlights, automatic | Rendered diffs are commit-relative, not read-relative |
| Agent integrations | PMD Works with Claude Code, Codex, Cursor, opencode | No cross-agent review loop in the preview |
| Folder-as-reading-list | PMD .md-only sidebar, tabs, search, no configuration | Explorer lists everything; one preview pane per file |
| Rendered diff view | Read-relative change highlights | VSC Rendered git diffs with word-level emphasis (official docs, checked Sept 2026) |
| Price | One payment of $19 after the trial week | VSC Free and open source |
| Editing & extensions | Read-only by design; the agent edits | VSC Full editor plus the marketplace |
| Linux | Not yet | VSC Yes |
| Mermaid diagrams | Yes | Yes, built in since v1.121 |
| GitHub Flavored Markdown | Yes, full spec | CommonMark via markdown-it; Microsoft’s FAQ says GFM is not a target |
1An objection you can point at
The comment is a first-class object in the document. Highlight step 3, write “this assumes we can batch the migration; we can’t,” and the note sits exactly where the wrong assumption lives, in a plain sidecar file your agent ingests on its next pass. The agent replies in the same thread and updates the Markdown. You never translate document locations into prose for a chatbot.
The preview has no annotation surface of any kind. Every piece of feedback becomes a message somewhere else (Claude Code’s terminal, the Cursor chat, whatever your loop is), and you spend sentences re-establishing context the document already contained.
2“What moved since Tuesday?”
Change tracking is keyed to your reading, not to git. Open a file you’ve read before and the lines the agent rewrote are marked. Review shrinks from “re-read the whole plan” to “read the deltas.”
The diff-view preview is excellent at “what changed in this commit.” Uncommitted agent output, the normal case mid-loop, has no such affordance, so the honest answer the preview gives is: re-read everything.
3The reading hour
Point the app at the project folder and every .md file arrives as a reading surface: outline, tabs, search, light or dark, focus mode when the spec is long. Nothing to configure; the app is read-only, so it also can’t mangle the file you’re reviewing.
You can make the preview the default editor for .md via workbench.editorAssociations, and it becomes a respectable reading view. It remains, though, a mode of a writing tool: the chrome around it is editor chrome, and your attention pays rent to it.
And the rounds VS Code takes cleanly
Free, forever, on every platform including Linux. The deepest Markdown tooling ecosystem in existence, from markdownlint to export pipelines. First-class rendered diffs and native Mermaid in a preview Microsoft actively maintains (the v1.121 built-in Mermaid support shipped recently enough that most listicles still call it an extension). And the preview is already installed on machines where PrettyMD would be a new download. For writing and editing, that list is decisive.
Pick by the shape of your week
PrettyMD earns its $19 if…
- Your agents draft the specs and READMEs, and you are the reviewer
- Your feedback currently lives in chat messages you have to re-explain
- Revised documents should show you the deltas, not the whole file again
- You want a calm reading surface instead of one more editor pane
Stay in VS Code if…
- You write the Markdown yourself and want linting and extensions around it
- The budget is zero, non-negotiably
- Linux is your daily driver
- One tool for everything beats the right tool for one thing
Straight answers
Is PrettyMD a real alternative to the VS Code Markdown preview?
For a specific job, yes. If agents write the Markdown and you review it, PrettyMD gives you read-relative change tracking and line comments a compatible agent reads and acts on, two things the preview does not attempt. If you mainly write Markdown yourself, the preview is the better and free choice.
Does the VS Code Markdown preview render GitHub Flavored Markdown?
Not as a target. Microsoft’s own documentation answers “No” on GFM and states the preview follows the CommonMark specification via markdown-it (verified against code.visualstudio.com, September 2026). PrettyMD renders the full Markdown spec, Mermaid included.
Can I leave comments for my AI coding agent inside VS Code?
The Markdown preview has no annotation mechanism. Feedback happens in your agent’s chat or terminal, re-typed by you. PrettyMD’s comment loop (highlight, note, sidecar JSON, agent reply) exists precisely to remove that re-typing.
Does VS Code highlight what changed in a Markdown file?
Relative to a commit, yes: the preview can render diffs with word-level highlights. Relative to your last read, the question a reviewer actually asks about uncommitted agent output, no tool in VS Code answers it. That’s PrettyMD’s change tracking.
Is the $19 a subscription?
No. $19 one-time after a 7-day free trial, no card required for the trial, no account required at all.
Fact-checked September 9, 2026 against code.visualstudio.com/docs/languages/markdown (CommonMark/markdown-it statement, rendered diff view, built-in Mermaid since v1.121, editor-association setting) and code.visualstudio.com/download (free, open source). Visual Studio Code is a trademark of Microsoft Corporation; no affiliation or endorsement connects this page to Microsoft.