Markland Quickstart
Publish a markdown doc with your agent in five steps. Under two minutes, start to share link.
Markland is an MCP-based document publishing platform. Your AI agent calls a tool, a doc goes up, a share link comes back. This quickstart takes you from zero to your first shared document using Claude Code. The same steps work with any other MCP-compatible client.
Before you start
You'll need:
- A terminal with
claude(the Claude Code CLI) installed and signed in. - An email address you can receive a magic link on.
- About two minutes.
No credit card. No API keys to copy. Markland issues you a scoped bearer token the first time you connect Claude Code, and the CLI stores it for you.
1. Sign up
Open the Markland landing page, enter your email, and click the magic link that lands in your inbox. That's the whole signup flow — no password to set, no profile to fill out.
Once you're signed in you'll see the Explore page. Keep that tab open; you'll come back to it in step 5 to see your doc.
2. Wire up the MCP server
Open Claude Code (CLI or web) and send this message:
Install the Markland MCP server from https://markland.dev/setup
Claude Code fetches the runbook at that URL, walks you through a one-time browser authorization, and then
runs claude mcp add with the right transport flag and bearer token. When it's done,
claude mcp list should show markland in the list.
Under the hood Markland uses the Model Context Protocol — the tool surface your agent actually calls. There is no REST SDK to learn; the tool names are the API.
3. Publish your first doc
Ask Claude:
Publish a markdown doc titled “Hello Markland” with some notes about my project.
Your agent calls the markland_publish tool. You'll get back a share URL that looks like
/d/<share-token>. The doc belongs to you, is private by default, and only you (and any agents you
authorize on your behalf) can read it.
You can ask Claude to publish any markdown — a spec, a plan, a research note, a CLAUDE.md file, a
daily standup. Markland isn't opinionated about shape, only about being markdown.
4. Share it
Now grant access to a teammate:
Grant view access on that doc to friend@example.com.
Your agent calls markland_grant. Your friend receives an email with a link. If they're already signed in
to Markland they see the doc right away; if not, they click the magic link in the invite email and are dropped onto
the doc after signing in.
Grants are per-doc. You can grant view, comment, or edit. You can also grant to another agent by its agent ID, so an agent on a teammate's laptop can read and append to docs on yours.
5. View the doc
Open the share link in your browser. You'll see the rendered markdown. Head back to /explore and you'll see the doc listed under Mine + Shared.
Troubleshooting
-
claude mcp listdoesn't show markland. Re-paste “Install the Markland MCP server from https://markland.dev/setup” into Claude Code — the runbook is idempotent and safe to re-run; it'll allocate a fresh authorization and re-register the server. -
The agent says it doesn't know about
markland_publish. Restart Claude Code. New MCP servers are only picked up at session start. - The share link returns 404. The token is case-sensitive and must include every character. Paste the whole thing rather than retyping.
-
I want to revoke a grant.
Ask Claude: “Revoke access for friend@example.com on that doc.”
The agent calls
markland_revoke.
What's next
Browse public docs to see how other people are using Markland. Read the alternatives page if you're weighing this against Git, Notion, Google Docs, or HackMD. Stuck on anything the troubleshooting list doesn't cover? Reply to the signup email — a real human reads it.
Frequently asked questions
Do I need an Anthropic account?
No. You need a Markland account (created via magic link sent to your email) and an MCP-compatible client. Claude Code is the example throughout this quickstart because the claude mcp add CLI is the cleanest setup, but any client that registers an MCP server works.
Does Markland work with Cursor, Continue, or other MCP clients?
Yes. Markland exposes a standard MCP server. Any client that can register one — Cursor, Continue, Codex, Claude Desktop, custom agents — can call the eight Markland tools. The Claude Code paste-the-runbook flow above is Claude-Code-specific; for other clients, sign in to Markland, visit /settings/tokens to mint a personal token, then point your client's MCP config at https://markland.dev/mcp/ with an Authorization: Bearer mk_usr_… header. The MCP config.json snippet on the home page is the canonical form.
Where do my published docs live?
On Markland's application server, in SQLite, currently hosted on Fly.io in the US-East region. Documents are bytes-on-disk — what your agent writes is what readers see. See /security for hosting and encryption details.
What if I want to revoke an agent token?
Sign in to Markland, open /settings/agents (or /settings/tokens if you authenticated via personal token), and click revoke. Existing share-token URLs for documents you published continue to work; the revoked agent token just can't issue new MCP calls.
What does markland_publish actually do?
It creates a new document with the markdown body you pass in and returns its stable ID plus a share URL. Subsequent edits go through markland_update with an if_version argument so concurrent writers see a clean conflict instead of silently clobbering each other.
Can I make a doc publicly visible without granting access to individuals?
Yes — pass public=true to markland_publish (or markland_update). The doc appears on /explore and is reachable by anyone with the share URL, no grant required. Default is private.
Don’t have an account yet?
Drop your email and we’ll send a magic link the moment a slot opens. No password, no pitch deck, no spam.