--- title: "mdfy.cc — Interim Website Update & Business Direction" url: https://mdfy.app/f3f32npB updated: 2026-05-03T12:41:41.190Z source: "desktop" --- # mdfy.cc — Interim Website Update & Business Direction > Integrated work document for Claude Code handover. This single document provides all decisions, copy, manifesto, and implementation plans for self-sufficiency. > > Last updated: 2026-04-27 Launch deadline: 2026-06-16 (Week 7 Tuesday, HN Show HN) --- ## 0. Work Overview ### Purpose Create an **interim website state** that simultaneously showcases the future vision (Memory Layer + Bundle + Standard) without negating the current build status of mdfy.cc. ### Two Audiences 1. **Users**: Can try it immediately, clear Phase 1 value. 2. **Investors**: Clear future vision + roadmap + ambition. ### Scope of Work - Update English site (About, Manifesto, Pricing, MCP docs) - Create new Korean site (mdfy.cc/ko/) - Consistency across all pages + cross-linking ### Implementation Timeline - Content writing: This week - Site push: Next week - Polish: The following week --- ## 1. Founder Context - **Build Period**: 1 month as a side project - **Current Status**: Full-time mdfy commit, 0 users, 0 marketing - **Resource Constraints**: None (no capital/time pressure) - **Ambition**: Standard-setting movement - **Founder**: Hyunsang at Raymind.AI --- ## 2. Core Brand Message ### Manifesto (One-line definition) > **Own your markdown. Use it anywhere.** ### Two Layer Framing **Layer 1: A markdown tool for all users** - Capture AI answers, WYSIWYG edit, permanent URL - Valuable even without Memory features - Alternative to HackMD, GitHub Gist - Immediate utility **Layer 2: A memory layer for power users** - Bundle, semantic search, MCP, API - Curated memory authored by the user - AI agent integration - Long-term moat + differentiation Key: **Both layers are valuable. Memory is a natural evolution, not a forced requirement.** ### 5 Beliefs (Manifesto core) 1. **Markdown is the right primitive for AI-era knowledge.** 2. **URLs are the right interface.** 3. **Memory is yours, not extracted.** 4. **Memory should be deployable.** 5. **Open by default.** --- ## 3. Hero Copy (Confirmed) ### English Hero **H1**: Own your markdown. Use it anywhere. **Sub**: Capture AI answers. Edit in WYSIWYG. Share, bundle, deploy as context. Owned, edited, portable. **CTA primary**: Start writing → **CTA secondary**: Install Chrome extension ### Korean Hero **H1**: Your markdown, anywhere. **Sub**: ChatGPT and Claude answers in one place. Edit in markdown, use as context for any AI. Created by me, used by me, owned by me. **CTA primary**: Start now → **CTA secondary**: Install Chrome extension --- ## 4. Three Pillars (Capture / Edit / Use) ### Pillar 1: Capture **Headline (EN)**: Capture from any AI, anywhere. **Headline (KR)**: Capture from any AI, anywhere. **Sub (EN)**: From your AI chats, your editor, your terminal — into one URL. **Sub (KR)**: AI chats, editors, terminals — all into a single URL. **Body (EN)**: You’re talking to AI all day. Great answers come, then disappear into chat history you’ll never search again. Or you copy-paste them into Notion, lose context, and forget why they mattered. mdfy captures them where you are — in one click, with source attribution, ready to edit. **Body (KR)**: You talk to AI all day. Great answers appear, then vanish into chat histories you'll never look at again. Or you copy-paste them into Notion, losing context and forgetting why they were important. mdfy captures them right where you are — with one click, including source information, ready for editing. **Surfaces**: - 🌐 **Chrome Extension** — One click on any AI chat (ChatGPT, Claude, Gemini) - 💻 **VS Code** — Save markdown directly from your editor - 🖥️ **Mac App** — Clipboard watch, instant capture - ⌨️ **CLI** — `cat README.md | mdfy` from your terminal - 📋 **Paste** — Drop anything into mdfy.cc ### Pillar 2: Edit **Headline (EN)**: Edit in beautiful markdown WYSIWYG. **Headline (KR)**: Beautiful markdown WYSIWYG editing. **Sub (EN)**: Your voice. Your structure. Your truth. **Sub (KR)**: Your voice. Your structure. Your truth. **Body (EN)**: Captured AI answers are starting points, not endings. Polish a thought. Add your context. Connect related ideas. mdfy gives you a markdown editor that feels like writing, not like coding. Tags, folders, and search keep your knowledge findable as it grows. AI tools help when you want them — never push when you don’t. **Body (KR)**: Captured AI answers are the beginning, not the end. Refine your thoughts, add context, and connect ideas. mdfy provides a markdown editor that feels like writing, not coding. Tags, folders, and search ensure your knowledge remains findable as it grows. AI tools assist only when you want them, never forcing themselves upon you. **Features**: - ✨ **WYSIWYG** — Beautiful markdown, no syntax friction - 🏷️ **Tags & folders** — Organize how you think - 🔍 **Smart search** — Find by keyword or meaning - 🤖 **AI tools** — Polish, summarize, translate (optional) - 📜 **Version history** — Never lose a thought ### Pillar 3: Use **Headline (EN)**: Use it anywhere — read, share, deploy. **Headline (KR)**: Anywhere — read, share, deploy. **Sub (EN)**: A permanent URL. A clipboard ready for any AI. A bundle to share with your team. **Sub (KR)**: A permanent URL. A clipboard ready to paste into any AI. A bundle to share with your team. **Body (EN)**: Every mdfy document gets a permanent URL. Send it to a friend. Embed in your blog. Paste back into Claude as context. Let your AI agent fetch it via MCP. When you’re ready to go deeper: **Bundle multiple URLs into a single deployable context.** Project briefs, agent definitions, knowledge collections — all become one URL your AI can read. This is where mdfy stops being a publisher and starts being your AI memory. **Body (KR)**: Every mdfy document receives a permanent URL. Send it to a friend. Embed it in your blog. Paste it into Claude as context. Let your AI agent fetch it via MCP. When you want to go deeper: **Bundle multiple URLs into a single deployable context.** Project briefs, agent definitions, knowledge collections — all become a single URL that AI can read. At this point, mdfy moves beyond being a publisher to become your AI memory. **Capabilities**: - 🔗 **Permanent URL** — Share anywhere, lasts forever - 🌍 **Public, unlisted, or private** — Your choice - 📦 **Bundle as context** *(Pro, coming Q2 2026)* — Multiple URLs as one deployable unit - 🔌 **MCP & API** *(Build, coming Q2 2026)* — AI agents read and write directly - 🧬 **Version snapshots** *(Build, coming Q2 2026)* — Lock a moment in time --- ## 5. Vision Section (To be added to About page) ### English **Section title**: Where mdfy is going **Subtitle**: Today: a markdown tool. Tomorrow: the memory layer for AI-native work. **Body**: mdfy started as a markdown publishing tool — and it remains a great one. But the bigger bet is that markdown URLs become the substrate for AI-era knowledge. Here’s what’s already shipped, and what’s coming next. #### Today (Phase 1 — Live) Already built and working: - Multi-surface capture (Chrome, VS Code, Mac, CLI, MCP server) - Beautiful WYSIWYG editing - Permanent URLs with sharing - AI tools (polish, summary, translate, chat) - 14+ file format imports (PDF, DOCX, PPTX, etc.) - KaTeX math, Mermaid diagrams, syntax highlighting #### Tomorrow (Phase 2 — Coming Q2 2026) \[Coming Soon Label\] Memory layer features in active development: - **Memory Bundle** — Multiple mdfy URLs as a single deployable AI context. Capture your project’s full context — spec, decisions, notes, references — and paste it into any AI as one URL. - **Semantic search** — Find by meaning, not just keyword. “What did Claude tell me about LLM memory?” finds the right answers across your entire collection. - **Bundle versioning** — Snapshot a moment. Compare evolutions. Roll back when needed. Memory with a time dimension. - **MCP write access** — Let AI agents not just read, but write back to your memory. The full read-write loop for agentic workflows. #### Beyond (Phase 3+) \[Vision Label\] The bigger bet: - **mdfy Bundle Spec** — An open standard for AI memory bundles, published for any tool to implement. - **Bundle marketplace** — Share, fork, subscribe to curated knowledge bundles. - **Team workspaces** — Shared memory across organizations. - **Enterprise self-host** — Full control for organizations that need it. [Read the full manifesto →](/manifesto) ### Korean **Section title**: Where mdfy is headed **Subtitle**: Today: A markdown tool. Tomorrow: The memory layer for AI-native work. **Body**: mdfy started as a markdown publishing tool, and it remains an excellent one. However, the bigger bet is that markdown URLs will become the substrate for knowledge in the AI era. Take a look at what has already been shipped and what is coming next. #### Today (Phase 1 — Live) Already built and operational: - Multi-surface capture (Chrome, VS Code, Mac, CLI, MCP server) - Beautiful WYSIWYG editing - Permanent URLs and sharing - AI tools (polish, summary, translate, chat) - 14+ file format imports (PDF, DOCX, PPTX, etc.) - KaTeX formulas, Mermaid diagrams, code highlighting #### Tomorrow (Phase 2 — Coming Q2 2026) \[Coming Soon Label\] Memory layer features under active development: - **Memory Bundle** — Multiple mdfy URLs as a single deployable AI context. Capture the entire context of a project—specs, decisions, notes, references—and paste it into any AI as a single URL. - **Semantic Search** — Find by meaning, not just keywords. "What did Claude say about LLM memory?" Discover exact answers across your entire collection. - **Bundle Versioning** — Snapshot a moment. Compare evolutions. Roll back if necessary. Memory with a temporal dimension. - **MCP Write Access** — Allow AI agents not just to read, but to write to memory. A full read-write loop for agentic workflows. #### Beyond (Phase 3+) \[Vision Label\] The bigger bet: - **mdfy Bundle Spec** — An open standard for AI memory bundles, released so any tool can implement it. - **Bundle Marketplace** — Share, fork, and subscribe to curated knowledge bundles. - **Team Workspaces** — Shared memory across an entire organization. - **Enterprise Self-host** — For organizations requiring complete control. [Read the full manifesto →](/ko/manifesto) --- ## 6. Comparison Table Addition (vs AI Memory Solutions) ### English **Section title**: mdfy vs AI Memory Solutions **Subtitle**: A different approach to AI memory. Authored by you, not extracted by AI. **Table**: | | mdfy.cc | Mem0 | Letta | Notion AI | | --- | --- | --- | --- | --- | | Author your own memory | ✓ | — | — | △ | | Markdown native | ✓ | — | — | △ | | Permanent URLs | ✓ | — | — | — | | Multi-LLM agnostic | ✓ | ✓ | ✓ | — | | Open source engine | ✓ | ✓ | ✓ | — | | Bundle for deployment | ✓ | — | — | — | | Human-readable storage | ✓ | △ | △ | ✓ | | MCP support | ✓ | — | — | — | | WYSIWYG editing | ✓ | — | — | ✓ | | No vendor lock-in | ✓ | ✓ | ✓ | — | **Note (below table)**: > Mem0 and Letta are excellent at what they do — they extract memory from your AI conversations automatically. > > mdfy answers a different question: what do *you* want to remember? You author. You bundle. You deploy. ### Korean **Section title**: mdfy vs AI Memory Solutions **Subtitle**: A different approach to AI memory. Authored by you, not extracted by AI. **Table** (Same structure as English above) **Note**: > Mem0 and Letta are excellent in their respective fields — they automatically extract memory from AI conversations. > > mdfy answers a different question: What do *you* want to remember? You author it. You bundle it. You deploy it. --- ## 7. Pricing Card Update Maintain existing pricing structure but add Phase 2 features to the “Pro (AFTER BETA)” card: ### English — Pro Card ```text ### Pro AFTER BETA Pricing announced when beta ends * + Everything in Beta * + No badge on shared docs * + Custom domain * + View analytics * + Password protection * + Priority AI mdfy * + Tags, folders, semantic search [Coming Q2 2026] * + 5 Memory Bundles [Coming Q2 2026] * + Version history ``` ### English — New Tier “Build” Card ```text ### Build COMING Q2 2026 [Coming Soon Label] For AI builders and power users * + Everything in Pro * + Unlimited Memory Bundles * + API access (read/write) * + MCP server (full, write enabled) * + Bundle versioning + snapshots * + Bundle analytics * + Webhook integrations * + Format adapters (Claude XML, etc.) ``` ### Korean — Pro Card ```text ### Pro After Beta Pricing announced when beta ends * + Everything in Beta * + Remove badge from shared docs * + Custom domain * + View analytics * + Password protection * + Priority AI mdfy * + Tags, folders, semantic search [Coming Q2 2026] * + 5 Memory Bundles [Coming Q2 2026] * + Version history ``` ### Korean — Build Card ```text ### Build Coming Q2 2026 [Coming Soon Label] For AI builders and power users * + Everything in Pro * + Unlimited Memory Bundles * + API access (read/write) * + MCP server (full, write enabled) * + Bundle versioning + snapshots * + Bundle analytics * + Webhook integrations * + Format adapters (Claude XML, etc.) ``` --- ## 8. Manifesto Page (NEW — mdfy.cc/manifesto) ### English Manifesto (Full) # Why I’m building mdfy I built mdfy in one month while doing other work. Now I’m going full-time. This is why. ## The state of AI memory today Every day, millions of people pour their thinking into ChatGPT, Claude, Gemini, and Cursor. We ask hard questions. We get back genuinely useful answers — strategies, code, frameworks, insights that took experts decades to develop. Then we close the tab. That answer is gone. Not literally — it sits in some chat history we’ll never search. But functionally gone. We can’t find it. We can’t reuse it. We can’t build on it. The next day, we ask similar questions. We get similar answers. We close the tab again. This is happening at civilizational scale. Trillions of tokens of high-quality, AI-assisted thinking, evaporating into chat histories nobody returns to. The world’s most expensive forgetting machine. The industry’s response so far is what I call **extracted memory** — services like Mem0 and Letta that watch your conversations and extract facts the AI thinks are important. “Sarah is vegetarian. Sarah lives in Seoul. Sarah is interested in LLM evaluation.” Mem0 and Letta are excellent at what they do. They solve a real problem. But they answer a different question than the one I want to answer. They ask: *what should the AI remember about you?* I want to ask: *what do you want to remember?* These are not the same question. The first is about inference. The second is about authorship. ## Why authorship matters Memory is not just data. Memory is identity. What you remember shapes who you become. What an organization remembers shapes what it can do. This was true in the age of paper, true in the age of databases, and is more true in the age of AI than ever before. When you let an AI extract your memory, you let an AI define what mattered. You let an algorithm decide which thread of yesterday’s thinking is worth carrying forward, which insight to compress into a fact, which piece of yourself to keep. That’s a strange thing to outsource. Some people will outsource it gladly. The convenience is real. But for those of us who think carefully about what we want our future selves to know — for those of us who treat our knowledge as a craft, not a byproduct — there should be another option. That option is mdfy. ## What mdfy is today If you visited mdfy.cc right now, you’d see what looks like a markdown publishing tool — and it is. You can capture markdown from anywhere: ChatGPT, Claude, Gemini (via Chrome extension), GitHub repos, your terminal (`cat README.md | mdfy`), VS Code, your Mac clipboard. You can edit it in a beautiful WYSIWYG editor — no syntax friction, no install required. You can share it with a permanent URL that anyone can read in the browser, that any AI can fetch as context. It’s a publishing tool. It works. People can use it today. In one month — built nights and weekends — I shipped: - A Rust markdown engine (mdcore, open source) - A web editor with WYSIWYG - A Chrome extension for any AI chat - A VS Code extension - A Mac desktop app - A CLI - An MCP server I shipped this fast because I had a clear primitive: the markdown URL. Every surface points to the same thing. Every surface composes with the others. That’s where mdfy is now. A useful markdown tool with a clean primitive. That’s not where it’s going. ## The bigger bet The bigger bet is that **markdown URLs are the right substrate for AI-era knowledge**. Not as a publishing tool. As infrastructure. Here’s the thesis: LLMs read and write markdown natively. It’s the lingua franca they were trained on. When ChatGPT outputs structured information, it outputs markdown. When you paste context into Claude, you paste markdown. When agents communicate with each other, the natural format is markdown. This is not changing. It’s compounding. Humans also read markdown natively. Plain text formatted lightly is how we’ve taken notes for centuries. It’s how we’ll keep taking them. No proprietary format will displace it. URLs are the simplest possible interface. Anyone can paste them. Any agent can fetch them. They cross every boundary — operating systems, applications, AIs, time zones, decades. If LLMs write markdown, humans read markdown, and URLs cross every boundary, then the natural primitive for AI-era knowledge is **a markdown document at a URL**. This is what mdfy is. This is what mdfy will become more deeply. ## What’s coming next The next eight weeks of building are about turning mdfy from a publishing tool into a memory layer. The features have names — Memory Bundle, Semantic Search, Bundle Versioning — but the underlying idea is one thing: **You should be able to take what you’ve authored and deploy it as context to any AI, anywhere.** The Memory Bundle is the deployment unit. Take five mdfy URLs that together describe your project — the spec, the design decisions, the recent meeting notes, the customer interview, the open questions — and bundle them into a single URL. Paste that one URL into Cursor, Claude, ChatGPT. Your AI now has the full context, in your words, organized your way. Bundle versioning lets you snapshot moments. The spec as it was when the project started. The spec as it is now. The diff between them, inspectable by you and by the AI you’re asking for help. Semantic search lets you find by meaning, not just keyword. “What did Claude tell me about LLM memory architecture?” returns results even if the words don’t match exactly. These three features turn mdfy from a place where you store markdown into a place where your AI memory lives — authored by you, deployable everywhere. ## The five beliefs Behind every product decision are beliefs. Here are mine. **1. Markdown is the right primitive for AI-era knowledge.** Not Notion blocks. Not Obsidian’s proprietary linking. Not closed app formats. Plain markdown. It’s what LLMs speak. It’s what humans read. It will outlast every tool currently fighting to capture knowledge. **2. URLs are the right interface.** Not SDKs. Not vendor lock-in. Not “install our app to access your data.” A URL — pastable, fetchable, openable in any browser, by any human, by any AI. The simplest possible interface is the most durable. **3. Memory is something you author, not something extracted.** Mem0 and Letta extract memory from your behavior. mdfy lets you write it, edit it, decide what stays. Both approaches are valid. One serves convenience. The other serves intention. mdfy is for people who want intention. **4. Memory should be deployable.** Storage isn’t the goal. Retrieval and reuse are. A memory you can’t paste back into an AI as context isn’t doing the work memory is supposed to do. Bundles make memory deployable. **5. Open by default.** mdcore — the engine — is open source. Markdown is an open standard. The Bundle spec will be published openly so other tools can implement it. Open formats and open interfaces are how durable infrastructure gets built. Closed systems are how vendors trap users. We choose open. ## Why now There’s a narrow window where this matters. For the past two years, the industry has been building closed AI memory systems — OpenAI Memory inside ChatGPT, Google’s Memory Bank inside Gemini, the various extracted-memory startups inside their own SDKs. Each one is trying to own your memory inside their walls. In another two years, one of two things will be true. Either the closed systems will have won, and memory will be something that lives inside whichever AI vendor you use — owned by them, queryable only through their API. Or an open standard will have emerged, and memory will be something you carry across vendors, in formats you control. I’m betting on the second outcome. I’m betting that markdown URLs become the open standard for AI memory the way HTTP became the open standard for documents — not because anyone declared it so, but because the primitive was right and the alternatives were worse. mdfy exists to make that outcome more likely. ## Why mdfy specifically Three things make mdfy different from anything else trying to do this. **Multi-surface from day one.** Most memory startups have a web app and an API. mdfy has a Chrome extension that captures from any AI, a VS Code extension, a Mac app, a CLI, and an MCP server — all built and shipped. Memory is only useful if it goes where you go. mdfy goes everywhere. **Markdown-native, not markdown-as-export.** Notion can export markdown but isn’t markdown. Obsidian uses markdown but locks it to a folder. mdfy treats markdown as the primary format, the URL as the primary identifier, and the WYSIWYG editor as a UI on top of that. Markdown isn’t a format mdfy supports — it’s what mdfy *is*. **Open from day one.** mdcore is open source. The Bundle spec will be open. Self-hosting will be available for enterprise. Trust requires that users can leave with their data, in formats other tools can read. We build for that from the start, not as a marketing afterthought. ## The roadmap **Phase 1 — now (live).** mdfy as a markdown publishing tool. Capture from any AI, edit in WYSIWYG, share with permanent URLs. Free during beta. **Phase 2 — next 8 weeks.** Memory Bundle, Semantic Search, Bundle Versioning. The transition from publishing tool to memory layer. Pro and Build tiers activate. **Phase 3 — Year 1.** MCP write access for AI agents. Bundle as context in production AI systems. Team workspaces with shared bundles. Open Bundle Spec v1.0 published for community review. **Phase 4 — Year 2-3.** Bundle marketplace. Enterprise self-host. Standard-setting consortium with other open-memory advocates. Acquihire conversations or Series A, depending on what serves the mission. I’m building this as if it could become infrastructure. Most of the time, that’s not what something becomes. But you can’t build infrastructure unless you build as if it could be. ## An open invitation If you’ve read this far, you probably care about one of these things: AI memory, markdown as a primitive, open standards, indie tools that don’t lock you in, or some combination. Here’s what I’m asking from you. **If you use AI daily**, try mdfy. The Chrome extension is the fastest entry. Capture a few good answers. See if you start coming back. The beta is free. **If you build AI agents or tools**, look at the MCP server. The current capabilities are read-only; write access is coming in Phase 2. If your product needs a memory layer, mdfy might be the right one — or it might be the wrong one, and I want to hear why. **If you care about open standards**, the Bundle spec is coming. I want feedback before it ships. The version that gets adopted will be better than the one I’d write alone. **If you’re an investor**, I’m not raising right now. I will be when the metrics justify it. If this thesis resonates, the conversation worth having is at that point — and I want to be talking to investors who care about open infrastructure, not closed unicorns. **If you just like reading about this stuff**, thank you for reading this far. The conversation around AI memory is going to define a lot of the next decade. The more people thinking about it carefully, the better. ## A note on what’s next I’m building in public. The next eight weeks will be visible — every feature shipped, every decision made, every pivot when something doesn’t work. Some of it will be ugly. Some of it will be wrong. All of it will be honest. The first launch — Memory Bundle, Semantic Search, public — is targeted for the end of June 2026. Until then: capture freely, share openly, and tell me what’s missing. --- *mdfy is built by Hyunsang at Raymind.AI.* *The mdcore engine is [open source on GitHub](https://github.com/raymindai/mdcore).* *The Bundle spec will be published before Phase 2 ships.* *Reach me at [hi@raymind.ai](mailto:hi@raymind.ai).* ### Korean Manifesto (Full — Re-creation) # Why I am building mdfy I built mdfy in one month while working on other things. Now, I am going full-time. This is why. ## The reality of AI memory today Every day, millions of people pour their thoughts into ChatGPT, Claude, Gemini, and Cursor. We ask difficult questions. We receive truly useful answers—strategies, code, frameworks, and insights that took experts decades to develop. Then we close the tab. That answer disappears. Not literally—it remains in some chat history we never search again—but functionally, it's gone. We can't find it, we can't reuse it, and we can't build upon it. The next day, we ask similar questions. We get similar answers. We close the tab again. This is happening on a civilizational scale. Trillions of tokens of high-quality, AI-assisted thinking are evaporating into chat histories that no one returns to. The world's most expensive forgetting machine. The industry's response so far is what I call **extracted memory**—services like Mem0 and Letta that watch your conversations and extract facts the AI deems important. "Sarah is a vegetarian. Sarah lives in Seoul. Sarah is interested in LLM evaluation." Mem0 and Letta are excellent in their domain. They solve a real problem. But they answer a different question than the one I want to address. They ask: *What should the AI remember about you?* I want to ask: *What do you want to remember?* These are not the same question. The first is about inference. The second is about authorship. ## Why authorship matters Memory is not just data. Memory is identity. What you remember determines who you become. What an organization remembers determines what it can achieve. This was true in the age of paper, true in the age of databases, and it is truer than ever in the age of AI. When you let an AI extract your memory, you are letting the AI define what was important. You are letting an algorithm decide which thread of yesterday's thoughts to carry forward, which insight to compress into a fact, and which part of yourself to keep. This is a strange thing to outsource. Some will outsource it gladly. The convenience is real. But for those of us who think carefully about what we want our future selves to know—those of us who treat our knowledge as a craft, not a byproduct—there must be another option. That option is mdfy. ## What mdfy is now If you visit mdfy.cc right now, it will look like a markdown publishing tool—and it is. You can capture markdown from anywhere: ChatGPT, Claude, Gemini (via Chrome extension), GitHub repos, the terminal (`cat README.md | mdfy`), VS Code, and the Mac clipboard. You can edit it in a beautiful WYSIWYG editor—without syntax friction or installation. You can share it with a permanent URL that anyone can read in a browser and any AI can fetch as context. It is a publishing tool. It works. People can use it today. In one month—built during nights and weekends—I shipped: - A Rust markdown engine (mdcore, open source) - A WYSIWYG web editor - A Chrome extension that works in any AI chat - A VS Code extension - A Mac desktop app - A CLI - An MCP server I was able to ship this quickly because I had a clear primitive: the markdown URL. Every surface points to the same thing. Every surface composes with the others. That is where mdfy stands today. A useful markdown tool with a clean primitive. But that is not where it is going. ## The bigger bet The bigger bet is that **markdown URLs are the right substrate for knowledge in the AI era**. Not as a publishing tool, but as infrastructure. This is the thesis: LLMs read and write markdown natively. It is the lingua franca they were trained on. When ChatGPT outputs structured information, it outputs markdown. When you paste context into Claude, you paste markdown. When agents communicate with each other, the natural format is markdown. This is not changing; it is compounding. Humans also read markdown natively. Lightly formatted plain text is how we have taken notes for centuries. It is how we will continue to take them. No proprietary format will replace this. URLs are the simplest possible interface. Anyone can paste them. Any agent can fetch them. They cross every boundary—operating systems, applications, AI, time zones, and decades. If LLMs write markdown, humans read markdown, and URLs cross every boundary, then the natural primitive for AI-era knowledge is a **markdown document at a URL**. This is what mdfy is, and this is what mdfy will more deeply become. ## What's coming next The next eight weeks of building are about transforming mdfy from a publishing tool into a memory layer. The features have names—Memory Bundle, Semantic Search, Bundle Versioning—but the underlying idea is one: **You should be able to deploy what you have authored as context to any AI, anywhere.** The Memory Bundle is the unit of deployment. Take five mdfy URLs that together describe a project—specs, design decisions, recent meeting notes, customer interviews, open questions—and bundle them into a single URL. Paste that one URL into Cursor, Claude, or ChatGPT. Your AI now has the full context, in your words, organized your way. Bundle versioning allows you to snapshot moments. The spec as it was when the project started. The spec as it is now. The diff between them, inspectable by both you and the AI you are asking for help. Semantic search allows you to find by meaning, not just keywords. "What did Claude say about LLM memory architecture?" returns results even if the words don't match exactly. These three features turn mdfy from a place where you store markdown into a place where your AI memory lives—authored by you and deployable everywhere. ## Five Beliefs Behind every product decision are beliefs. These are mine. **1. Markdown is the right primitive for AI-era knowledge.** Not Notion blocks. Not Obsidian's proprietary linking. Not closed app formats. Plain markdown. It's what LLMs speak. It's what humans read. It will outlast every tool currently fighting to capture knowledge. **2. URLs are the right interface.** Not SDKs. Not vendor lock-in. Not "install our app to access your data." A URL—pastable, fetchable, openable in any browser, by any human, by any AI. The simplest possible interface is the most durable. **3. Memory is something you author, not something extracted.** Mem0 and Letta extract memory from your behavior. mdfy lets you write it, edit it, and decide what stays. Both approaches are valid. One serves convenience; the other serves intention. mdfy is for people who want intention. **4. Memory should be deployable.** Storage is not the goal. Retrieval and reuse are. A memory you cannot paste back into an AI as context is not doing the work memory is supposed to do. Bundles make memory deployable. **5. Open by default.** mdcore—the engine—is open source. Markdown is an open standard. The Bundle spec will be published openly so other tools can implement it. Open formats and open interfaces are how durable infrastructure is built. Closed systems are how vendors trap users. We choose open. ## Why now? There is a narrow window where this matters. For the past two years, the industry has been building closed AI memory systems—OpenAI Memory inside ChatGPT, Google's Memory Bank inside Gemini, and various extracted-memory startups inside their own SDKs. Each is trying to own your memory within their walls. In another two years, one of two things will be true. Either the closed systems will have won, and memory will be something that lives inside whichever AI vendor you use—owned by them and queryable only through their API. Or an open standard will have emerged, and memory will be something you carry across vendors in formats you control. I am betting on the second outcome. I am betting that markdown URLs will become the open standard for AI memory the way HTTP became the open standard for documents—not because anyone declared it so, but because the primitive was right and the alternatives were worse. mdfy exists to make that outcome more likely. ## Why mdfy specifically? Three things make mdfy different from anything else trying to do this. **Multi-surface from day one.** Most memory startups have a web app and an API. mdfy has a Chrome extension that captures from any AI, a VS Code extension, a Mac app, a CLI, and an MCP server—all built and shipped. Memory is only useful if it goes where you go. mdfy goes everywhere. **Markdown-native, not markdown-as-export.** Notion can export markdown but is not markdown. Obsidian uses markdown but locks it to a folder. mdfy treats markdown as the primary format, the URL as the primary identifier, and the WYSIWYG editor as a UI on top of that. Markdown is not a format mdfy supports; it is what mdfy *is*. **Open from day one.** mdcore is open source. The Bundle spec will be open. Self-hosting will be available for enterprise. Trust requires that users can leave with their data in formats other tools can read. We build for that from the start, not as a marketing afterthought. ## Roadmap **Phase 1 — Now (Live).** mdfy as a markdown publishing tool. Capture from any AI, edit in WYSIWYG, share with permanent URLs. Free during beta. **Phase 2 — Next 8 weeks.** Memory Bundle, Semantic Search, Bundle Versioning. The transition from publishing tool to memory layer. Pro and Build tiers activate. **Phase 3 — Year 1.** MCP write access for AI agents. Bundles as context in production AI systems. Team workspaces with shared bundles. Open Bundle Spec v1.0 released for community review. **Phase 4 — Year 2-3.** Bundle marketplace. Enterprise self-host. Standard-setting consortium with other open-memory advocates. Acquihire conversations or Series A, depending on what serves the mission. I am building this as if it could become infrastructure. Most of the time, things don't become that. But you cannot build infrastructure unless you build as if it could be. ## An open invitation If you've read this far, you probably care about one of these things: AI memory, markdown as a primitive, open standards, indie tools that don't lock you in, or some combination. Here is what I am asking of you. **If you use AI daily**, try mdfy. The Chrome extension is the fastest entry point. Capture a few good answers. See if you start coming back. The beta is free. **If you build AI agents or tools**, look at the MCP server. Current capabilities are read-only; write access is coming in Phase 2. If your product needs a memory layer, mdfy might be the right one—or it might be the wrong one, and I want to hear why. **If you care about open standards**, the Bundle spec is coming. I want feedback before it ships. The version that gets adopted will be better than the one I would write alone. **If you are an investor**, I am not raising right now. I will be when the metrics justify it. If this thesis resonates, the conversation worth having is at that point—and I want to be talking to investors who care about open infrastructure, not closed unicorns. **If you just like reading about this stuff**, thank you for reading this far. The conversation around AI memory will define much of the next decade. The more people thinking about it carefully, the better. ## A note on what's next I am building in public. The next eight weeks will be visible—every feature shipped, every decision made, every pivot when something doesn't work. Some of it will be ugly. Some of it will be wrong. All of it will be honest. The first launch—Memory Bundle, Semantic Search, public—is targeted for late June 2026. Until then: capture freely, share openly, and tell me what's missing. --- *mdfy is built by Hyunsang at Raymind.AI.* *The mdcore engine is [open source on GitHub](https://github.com/raymindai/mdcore).* *The Bundle spec will be published before Phase 2 ships.* *Reach me at [hi@raymind.ai](mailto:hi@raymind.ai).* --- ## 9. Site Structure — All Pages ### English Site ```text mdfy.cc/ Home (Editor) mdfy.cc/about About (Main Marketing) mdfy.cc/manifesto Manifesto (NEW) mdfy.cc/plugins Plugins (Minor update) mdfy.cc/docs Docs landing mdfy.cc/docs/api REST API mdfy.cc/docs/cli CLI mdfy.cc/docs/sdk SDK mdfy.cc/docs/mcp MCP Server (Minor update) mdfy.cc/privacy Privacy ``` ### Korean Site (NEW) ```text mdfy.cc/ko/ Home (Link to English editor or Korean landing) mdfy.cc/ko/about About Korean mdfy.cc/ko/manifesto Manifesto Korean ``` **Note**: The editor itself will not be translated into Korean (English OK). --- ## 10. Page-by-page Update Specifications ### 10.1 mdfy.cc/about (Main Marketing) **Maintain existing structure but add/change**: #### Change 1: Hero - Existing: “The Markdown Hub — Collect. Edit. Publish.” - Change: New Hero from Section 3 above (English) - Keep existing hero image (`hero-editor.webp`) #### Change 2: Three Pillars Section (Collect / Edit / Publish) - Maintain 3-pillar structure - Update copy to Capture / Edit / Use from Section 4 above - Optional: Rename “Publish” to “Use” #### Change 3: Add Vision Section (NEW) - Use text from Section 5 above - Location: After “What it does” section, before “Comparison table 1” #### Change 4: Add Comparison Table (NEW) - Use text from Section 6 above (vs AI Memory Solutions) - Location: After existing table 1 (vs Markdown publishing tools) #### Change 5: Pricing Section - Use text from Section 7 above - Update “Pro” card + add “Build” card #### Change 6: Add Manifesto Link - Small section above “Try it now”: ```text ## The bigger picture This is more than a markdown tool. Read why I'm building mdfy. [Read the manifesto →](/manifesto) ``` ### 10.2 mdfy.cc/manifesto (NEW Page) Use English manifesto from Section 8 above. **Design Guidelines**: - Simple long-form essay layout - Center-aligned, readable width (\~720px) - Ample spacing between sections - Styling for blockquotes/emphasis - Visually distinguish the 4 audiences in “An open invitation” (cards or indentation) ### 10.3 mdfy.cc/ (Home/Editor) **Minimal changes**: - Maintain immediate editor opening - Add small links to Header: “About” + “Manifesto” + “한국어” **Optional Addition** (Hyunsang's choice): - Small banner above Editor: “🚀 Phase 2: Memory Bundle coming Q2 2026 \[See roadmap →\]” ### 10.4 mdfy.cc/plugins (Minor update) Add Phase 2 implications to each plugin card (optional): - “Coming in Phase 2: Bundle support, MCP write access” ### 10.5 mdfy.cc/docs/mcp (Strategic Update) **Maintain existing docs but add to top of page**: ```text # MCP Server > The MCP-native memory layer for AI agents. > Read mdfy URLs as context today. Write memory back via MCP in Phase 2. ## Today (Available) - Read mdfy URLs as AI context - Document fetch via MCP - Auto-source detection ## Coming Q2 2026 [Coming Soon Label] - Memory write access - Bundle deploy - Multi-agent memory sharing - Real-time bundle sync ``` Followed by existing technical documentation. ### 10.6 mdfy.cc/ko/about (Korean Main) Maintain English about page structure but translate all text into Korean. **Copy to use**: - Hero: Korean hero from Section 3 - Three pillars: Korean body from Section 4 - Vision section: Korean body from Section 5 - Comparison table: Korean body from Section 6 (English keywords in table OK, Korean descriptions) - Pricing: Korean body from Section 7 ### 10.7 mdfy.cc/ko/manifesto (Korean) Use Korean manifesto from Section 8 above. ### 10.8 mdfy.cc/ko/ (Korean Home) Option A (Simple): Redirect to English editor + preserve Korean nav Option B (Full): Korean landing page (Hero + brief description) + Editor entry button I recommend **Option A** (Time efficiency). --- ## 11. UI/UX Design Spec ### “Coming Soon” Label/Badge **Visual**: - Small colored box (badge style) - Color: Gold or Pastel Blue (less aggressive) - Text: “Coming Soon” or “Coming Q2 2026” **Application Locations**: - Phase 2 features in Vision section - Phase 2 features in Pro/Build pricing cards - Phase 2 capabilities in MCP docs - Next to any instance of the word “Coming” **Example CSS** (Reference): ```css .coming-soon-badge { display: inline-block; padding: 2px 8px; border-radius: 12px; background-color: #FEF3C7; color: #92400E; font-size: 11px; font-weight: 500; margin-left: 8px; } ``` ### “Vision” Label For Phase 3+ items. Signals a more distant future. ```css .vision-badge { background-color: #E0E7FF; color: #3730A3; } ``` ### Manifesto Page Design - Center-aligned, max-width 720px - Readable line-height (1.7+) - Section headings as h2, body as p - 5 beliefs as numbered list or card grid - Distinct visual treatment for the 4 audiences in “An open invitation” ### Korean/English Switch **Header**: - “EN | KO” or “한국어 | English” in top right - Link to the other language version of the current page - If no equivalent page exists (e.g., no Korean docs), link to English --- ## 12. Implementation Order ### Day 1-3: Finalize Content - ✅ Complete English Manifesto (Included in this doc) - ✅ Complete Korean Manifesto (Included in this doc) - ✅ Confirm all copy (Included in this doc) ### Day 4-5: Update About Page (English) - Change Hero - Update Three Pillars body (Capture/Edit/Use) - Add Vision section - Add AI Memory comparison table - Update Pricing section + add Build card - Add Manifesto link ### Day 6: Build Manifesto Page (English) - New route mdfy.cc/manifesto - Design: Center-aligned long-form essay - Anchor links (sections), reading time display (optional) ### Day 7: Build Korean Site - /ko/ subdirectory structure (i18n setup) - Build /ko/about, /ko/manifesto - EN/KO switch in Header ### Day 8: Update MCP Docs + Plugins - Add Memory layer messaging to top of mdfy.cc/docs/mcp - Add Phase 2 implications to mdfy.cc/plugins cards (optional) ### Day 9: Polish + Cross-link - Check cross-links on all pages - Consistency of “Coming Soon” labels - Mobile testing - Header/Footer consistency across all pages ### Day 10: Investor Lens Review - Review entire site from an investor's perspective - Review from a user's perspective - Note any deficiencies (for next task) --- ## 13. Key Decisions Summary | Decision | Answer | | --- | --- | | Manifesto | “Own your markdown. Use it anywhere.” | | Key Differentiation | Memory Bundle (Phase 2) | | English Hero | Own your markdown. Use it anywhere. | | Korean Hero | Your markdown, anywhere. | | Three Pillars | Capture / Edit / Use | | Korean Domain | mdfy.cc/ko/ (subdirectory) | | Editor in Korean | No (English OK) | | Pricing | Pro $9, Build $19 (after beta) | | Coming Soon Display | Label + soft date “Q2 2026” | | Comparison Addition | vs AI Memory Solutions (Mem0, Letta, Notion AI) | | Manifesto Tone | Personal essay, founder voice | | Phase 2 Feature Exposure | Name + 1-line description | | Investor Assets | Separate after this work (one-pager, deck) | | Launch Deadline | 2026-06-16 (Week 7 Tuesday) | --- ## 14. Next Steps (After this work) After completing interim website work: 1. **Start 8-week launch work** — Build Memory Bundle, Semantic search, Bundle versioning 2. **Build separate investor assets** — One-pager, pitch deck, demo video 3. **Plan recruitment for 50 beta testers** 4. **Draft Bundle Spec v1.0** — Community feedback before release 5. **Prepare for HN Show HN, Product Hunt** — Week 7 launch --- ## 15. Notes for Hyunsang ### How to use this document This single document allows for self-sufficient work in Claude Code: - Copy can be copied directly - Manifesto can be used as is - Korean copy can be used as is - Follow the implementation order ### Pending Decisions (Let me know if needed) Parts that may require additional decisions during implementation: - Whether to display Phase 2 banner on home (mdfy.cc/) - Three pillars section names: Keep existing “Collect / Edit / Publish” vs new “Capture / Edit / Use” - Korean home (Option A redirect vs Option B Korean landing) - “Coming Soon” color (Gold vs Blue, etc.) These decisions can be made naturally during implementation. ### If you need help If you get stuck: - Need additional copy (e.g., new page) - More detailed design specs - Alternative Korean expressions - Manifesto revisions You can start a conversation again at any time. --- *Document created: 2026-04-27* *Launch deadline: 2026-06-16* *Built by: Hyunsang at Raymind.AI* --- Result: