$mdflow use-cases --all
One workspace. Every reader.
MDflow is a folder of markdown files that people and AI agents read the same way. Here is what that is actually good for — each one a real capability, with what it does not do stated on its own page.
100 documents free, no card. Pro from €4.99/month.
$ls ./use-cases
Six things people do with it
Ordered roughly by how people arrive: the agent questions first, then the human ones.
- 01Share context across agentsClaude, Cursor and Codex read the same folders
- 02Let agents maintain a knowledgebaseThe agent files, you stay the editor
- 03Give automations a place to writeWebhooks in, documents out, share links back
- 04Share documents with othersA clean link anyone can open — no account, no install
- 05Save web pages as markdownClip the web, then let your agent read it too
- 06Keep sensitive notes encryptedEncrypted in your browser; the server stores ciphertext
$cat why.txt
What they have in common
Three properties every use case above inherits from the same design.
- One copy, many readers. The same folder is read by you in the editor, by a colleague through a share link, and by an agent over MCP. Nothing is exported, converted or duplicated to make that work.
- Markdown is the format, not an export. Every use case below operates on the same plain .md files. If you stop using MDflow tomorrow, what you take with you is what you already had.
- The folder description is the context. You write a sentence describing what belongs in a folder. That one sentence is the filing rule for agents, the ranking signal for retrieval, and a note to your future self.
$mdflow faq
Questions
Do I need Pro for these use cases?
Sharing documents and client-side encryption are free on every plan. The agent and automation use cases — the HTTP API, Personal Access Tokens and the MCP server — require Pro at €4.99/month, because they are what put load on the service.
Can people and AI agents read the same documents?
Yes, and that is the point. A document you write in the browser is readable by a colleague through a share link and by Claude, ChatGPT, Cursor or Codex over MCP, without any export step. Ownership is enforced server-side, so an agent only ever sees your workspace.
Which AI clients can connect to MDflow?
Claude.ai and the ChatGPT app connect over OAuth with a browser sign-in and no token to paste. Claude Code, Cursor, VS Code and the OpenAI Responses API use a Personal Access Token against the same hosted MCP server. Codex and other stdio-only clients can run the local server.
Can an AI agent write to my workspace, not just read it?
Yes. Agents can create, rename, move and rewrite documents and folders. Every automated write is recorded in the Document Log with the token's name, an edited row opens a diff of exactly what changed, and version history restores any previous version non-destructively.
Are encrypted documents readable by my agents?
No, deliberately. Encryption happens in your browser and MDflow only stores ciphertext, so an encrypted document is excluded from search, skipped by the Tasks scan and unreadable over MCP — including to your own agents.
$mdflow init
Start with one folder. Add readers later.
Write a document, describe the folder it lives in, then decide who reads it — a colleague on a link, an agent over MCP, or a workflow through the API.
100 documents free, no card. Sharing and encryption are free; the API, tokens and MCP are Pro from €4.99/month.