I use Claude probably 50 times a day. Not exaggerating. I think through problems with it, I plan features, I write specs, I draft things I need to send to people. A huge chunk of my actual work happens inside Claude conversations.
And for months, all of that work just… vanished. I’d have a great conversation on Tuesday, need it on Friday, and spend ten minutes scrolling through my history trying to find it. Half the time I’d give up and just re-explain everything from scratch. The AI didn’t remember. The conversation was gone. The work was gone.
That’s why I built MD Edit. Not because the world needed another text editor (it really didn’t) but because I needed a place for all this AI-generated work to actually live.
How I actually use it
Here’s my real workflow, every day:
I’m building MD Edit itself (yes, it’s recursive). I have a folder called “MD-Edit.com” with about 18 documents in it: tech specs, feature designs, this blog post you’re reading right now, SEO notes, go-to-market plans. All of them started as Claude conversations. All of them live in MD Edit now.
When I sit down to work on a new feature, I open a Claude conversation and say something like “read my tech spec for sync and versioning.” Claude reads the actual document, the latest version with all my edits, and I pick up right where I left off. No pasting. No re-explaining. It just has the context.
When I’m done, I tell Claude to save the output. Or I paste it into MD Edit myself. Either way, the work persists. It’s a real document I can find tomorrow.
The thing that surprised me is how much better Claude’s responses get when it has real context to work with. Starting every conversation cold is like hiring a contractor and refusing to show them the blueprints. Once Claude can actually read your previous work, the quality jumps noticeably.
The problem I kept running into
Before I built this, my workflow was embarrassing. I’d have Claude produce something great, like a detailed feature spec, a competitive analysis, or a plan I’d spent an hour refining. And then I’d… copy it into Apple Notes? Screenshot it? Leave it in the chat and hope I could find it later?
None of those worked. Apple Notes doesn’t understand markdown formatting, so everything Claude produces turns into a wall of plaintext. Google Docs works but feels heavy for this. I don’t need collaboration features and version history for a quick spec. And leaving things in Claude’s conversation history means you’re trusting a chat interface to be your filing system. It’s not designed for that.
The real kicker: you can’t feed your old work back into Claude. Even if you find that conversation from last Tuesday, you can’t say “hey Claude, read what we wrote last week and build on it.” You have to copy, paste, hope the formatting survives, and re-establish all the context manually.
I kept thinking: there has to be a place where I can write in markdown, have it look nice, save it permanently, organize it, and have Claude read it directly. Turns out there wasn’t. So I built one.
What it actually is
MD Edit is a markdown editor in your browser. You open md-edit.com, you start typing, and you have a document. No account. No signup. No “let me show you around” tour. Just an editor.
It does WYSIWYG, so you get bold, headings, lists, tables, code blocks, all rendered visually while you type. But it’s markdown underneath, so you can toggle to raw mode anytime. Everything AI tools produce is markdown, so it just works. Paste something from Claude or ChatGPT and it renders properly. Tables and all. If you’re weighing editors, I compared the best markdown WYSIWYG editors in a separate post.
Your documents live in your browser’s local storage. Nothing leaves your device unless you explicitly turn on sync. I’m pretty stubborn about this. I didn’t want to build something that watches what you write.
If you do want sync, there are two options: MD Edit’s own cloud (runs on Supabase), or your own Google Drive. The Google Drive sync goes directly from your browser to Google. I never see your files.
And then there’s the Claude connector. This is the part I’m most excited about. You add MD Edit as a connector in Claude’s settings, and suddenly Claude can read your documents, create new ones, search your library, and organize things into folders, all from inside a normal Claude conversation. You just talk to it naturally. “Save this as a doc called Feature Spec v2.” “Read my notes on the sync architecture.” “Move that to the Tech folder.” It works.
Prompts I actually use
Here are real things I say to Claude with the MD Edit connector. Not polished marketing examples, just stuff I actually type:
Saving work from a conversation:
- “Save this as a doc called ‘Sync Architecture v3’”
- “Take that competitive analysis and save it to my Research folder”
- “Create a new document with just the final version of that email”
Picking up where I left off:
- “Read my doc called ‘Q2 Roadmap’ and help me update the timeline”
- “Read the feature spec for sharing and tell me what edge cases I’m missing”
- “Search my documents for anything about pricing. I need to review what I’ve written before”
Organizing:
- “List everything in my MD-Edit.com folder”
- “Move the blog post draft to the Content folder”
- “What documents do I have about onboarding?”
Building on previous work:
- “Read my meeting notes from the investor call and draft a follow-up email”
- “Read my product spec and my competitor analysis, then write a positioning statement”
- “Read my last three blog post drafts and suggest what I should write about next”
The thing that makes this work is that Claude isn’t just reading a wall of pasted text. It’s reading a real document with a title, structure, and formatting intact. And because you can update the document and have Claude re-read it later, the context actually stays current.
My actual daily setup
I’ll walk you through what a real day looks like for me, because I think it’s more useful than a feature list.
Morning: I open Claude on my phone (Claude iOS app) while I’m having coffee. I think through whatever’s on my mind, usually a feature I want to build or a problem I’m stuck on. When the conversation gets somewhere useful, I tell Claude to save it to MD Edit. By the time I sit down at my computer, the document is already there.
Working: I open Claude Code in the terminal. I tell it to read my spec from MD Edit. It pulls up the latest version, including whatever I added from my phone that morning, and we start building. The spec is a living document. I update it as decisions get made. When something changes, I edit the doc directly in MD Edit and the next Claude session picks up the new version.
End of day: Whatever I produced that day (new specs, decision logs, notes on what worked and what didn’t) lives in my MD Edit library, organized by project. Tomorrow I’ll pick up exactly where I left off, without re-explaining anything.
The key thing: the AI gets better over time because it has real context. The first Claude conversation about a project starts cold. By the tenth conversation, Claude has read your spec, your decision log, your previous designs. The responses are just different: more specific, more aware of constraints, less generic. That only happens if the work persists somewhere Claude can access it.
What it’s not
I should be honest about what MD Edit doesn’t do, because I’d rather you know upfront.
It’s not Notion. No databases, no kanban boards, no team workspaces. I might build some of that eventually, but right now it’s a document editor. A really good one that works with your AI tools, but still a document editor.
It’s not a chat log backup. It doesn’t automatically save every Claude conversation. You decide what’s worth keeping. Most conversations aren’t. But the ones that are, the ones where real work got done, those deserve a permanent home.
It’s not trying to replace anything you already use. If Google Docs works for your team collaboration, keep using it. If Obsidian is your second brain, great. MD Edit fills a specific gap: the place where your AI work lives and compounds, accessible to Claude and ChatGPT and whatever comes next.
Why I’m writing this
Honestly, because I think a lot of people have the same frustration I did and don’t realize there’s a fix. You’re doing real work with AI. Meaningful, valuable work. And then you’re just… letting it disappear. Not because you’re lazy, but because there was nowhere good to put it.
There is now. I built it because I needed it. I use it every day. And it’s free. Open md-edit.com and start writing. That’s the whole pitch.
If you use Claude a lot, add the connector. It takes 30 seconds and it’ll change how you work with AI. Not in a dramatic, marketing-speak way. In a “oh, this is obviously how it should have worked the whole time” way.