Jonathon's AI Wiki

Simple Markdown Editor — Google Docs for Markdown, with Agents as First-Class Collaborators

6 min read

Source: ai-research/simplemarkdowneditor-landing-2026-07-03.md (https://simplemarkdowneditor.com, landing page as of 2026-07-03)

Simple Markdown Editor (simplemarkdowneditor.com) bills itself as “Google Docs, for markdown” — a browser-based collaborative editor with real-time multiplayer editing, margin comments, suggestions, and full version history. The reason it belongs in this wiki is its differentiator: a share link works for people and for agents — the same document, with the same comments, reachable over a CLI (mde cat <doc>) or plain HTTP. It’s the agent-readable-web thesis applied to collaborative document editing: humans and agents co-editing one markdown doc as equal participants. Start writing is instant, no account required, and each doc lives at its own private link.

Key Takeaways

  • “Google Docs for markdown.” Real-time collaborative editing over plain markdown files, positioned against the Google-Docs collaboration experience but keeping markdown as the native format (not a proprietary doc format).
  • CRDT multiplayer, no locks. Editing is backed by a CRDT (conflict-free replicated data type) — “no locks, no conflicts, multicursor included.” Multiple editors (human or agent) can type simultaneously without the check-out/merge-conflict model of file-based workflows.
  • Track-changes for markdown. “Suggest, don’t overwrite” — editors propose changes that are accepted or rejected with one click, rather than silently overwriting. This is the load-bearing feature for human-in-the-loop review of agent-proposed edits. ^[inferred — the landing page states track-changes/suggestions exist and separately states agents are first-class; that agent edits flow through the suggestion path is a reasonable but unstated combination]
  • Agents are first-class, not bolted on. The pitch explicitly includes “Made for agents too”: mde cat <doc> from a CLI, or hand any agent a share link. An agent reads and (per the collaboration model) edits the same doc a human sees, over the CLI or plain HTTP — no separate API surface, no scraping.
  • Zero-friction start, private-link model. No account needed to begin; the doc lives at its own private (unguessable) link. An optional username/password account exists but the core editor works without one.
  • Markdown-over-HTTP as the interface. Because a doc is reachable as plain markdown over HTTP, it needs no MCP server or SDK for an agent to consume it — a plain GET (or mde cat) is enough. This is a lighter-weight agent-access pattern than EmDash’s built-in MCP server. ^[inferred]

What it does (the collaboration layer)

  • Live multiplayer editing with shared cursors (see who’s editing where, in real time).
  • Comments and suggestions in the margin — the Google-Docs-style review surface, for markdown.
  • Full version history — every state of the doc is recoverable.
  • Instant, accountless docs — click “Start writing,” get a private link, share it. Optional account for (presumably) persistence/ownership across devices.

The agent angle (why it’s in this wiki)

The wiki tracks the shift to an agent-readable web — content authored so agents can read and act on it. Simple Markdown Editor is a supply/authoring-side instance of that shift, aimed at documents rather than websites:

  • One artifact, two audiences. The same share link is the human UI and the agent interface. Contrast the usual pattern where a human doc (Google Doc, Notion page) requires a separate API + auth for an agent to touch it.
  • CLI-native access. mde cat <doc> reads a doc’s markdown straight to stdout — the same ergonomics an agent already uses for local files (cat file.md), extended to a shared remote doc. This slots directly into a Claude Code / CLI-agent loop.
  • Review gate built in. Track-changes/suggestions give a human the accept/reject gate over agent-proposed edits — the same human-in-the-loop review pattern the wiki documents for agentic workflows, applied to prose/docs instead of code diffs. ^[inferred]

Nearest neighbors in the wiki: EmDash CMS (structured content emitted to agents via a built-in MCP server) and the WebMCP directory (sites exposing typed tools to agents). Simple Markdown Editor is the collaborative-document corner of that same space — and the lightest-weight, since its agent interface is just markdown over HTTP/CLI rather than a tool schema.

Implementation

  • Tool/Service: Simple Markdown Editor — https://simplemarkdowneditor.com (hosted web app; browser + mde CLI).
  • Setup: Open the site → “Start writing” → get a private doc link. No account required. For agent use, hand the agent the share link, or use mde cat <doc> from a shell. (The mde CLI’s install/distribution is not documented on the landing page — see Open Questions.)
  • Cost: Not stated on the landing page. The accountless free tier is confirmed; whether accounts or heavy use are paid is unknown. Treat pricing as data not available.
  • Integration notes: The agent-facing contract is “markdown over CLI/HTTP at a private link.” For a Claude Code loop, that means an agent can pull the current doc (mde cat), propose edits as suggestions, and let a human accept/reject — with no MCP server to stand up. CRDT backing means concurrent human+agent edits don’t need locking.

Try It

  1. Open a scratch doc. Go to https://simplemarkdowneditor.com/new — it opens instantly with a private link. Paste in some markdown.
  2. Test the collaboration. Open the same link in a second browser/profile; confirm real-time multicursor editing and margin comments.
  3. Test the agent path. Hand the share link to a CLI agent (or try mde cat <doc> if you have the CLI) and confirm it reads the current markdown. Then have the agent propose an edit and verify it lands as a suggestion you can accept/reject — not a silent overwrite.
  4. Compare to your current flow. If you presently round-trip docs to an agent by copy-pasting or syncing files, evaluate whether a single shared link (human + agent on one doc) removes steps.

Open Questions

  • Who builds/hosts it, and what’s the trust model? No maker attribution, company, or privacy/data-handling statement is on the landing page. For any sensitive content, the “private link” security model (unguessable URL) and data retention are unverified — treat as unknown before putting client/WEO-internal content in it.
  • Pricing. Not stated. Free accountless tier confirmed; paid tiers unknown.
  • The mde CLI. Install method, distribution (npm? brew? binary?), auth against private/owned docs, and whether it supports writes (not just cat) are undocumented.
  • Is there an MCP server or typed tool surface, or is markdown-over-HTTP the entire agent interface? The landing page implies the latter, but this isn’t confirmed.
  • Self-host / open-source? Unknown — no repo or license is linked.
  • Corroboration. No third-party reviews or coverage were found (web search returns unrelated same-named products). All claims here trace to the official landing page only; confidence is medium pending hands-on verification.
  • The Agent-Readable Web — the demand/interface/supply framing this tool is a supply-side instance of.
  • EmDash CMS — the closest neighbor: structured content emitted to agents via a built-in MCP server (heavier-weight than markdown-over-HTTP).
  • WebMCP Directory — sites exposing typed, callable tools to agents (the action-side counterpart).
  • Hardening an Agentic Prompt — the human-in-the-loop accept/reject review pattern, here applied to doc suggestions.
  • Firecrawl — LLM-ready content extraction; the read-the-web counterpart to authoring agent-readable docs.

Graph View

Backlinks