Skip to content
For Claude Cowork, Claude Code & every MCP agent.

Your AI agent made it. Now send the link.

Your AI agent writes the doc. AgentDoc gives it a real URL. Send it to anyone — no Claude account, no download, no copy-paste. Keep editing; the link stays live, with every edit named.

For coding agents & humans
One-click shareable links
Private by default, free
No Claude account to view

Your AI coworker made something great. Now you need to share it.

You ask your agent — Claude Code, Cowork, Cursor, Codex, whatever you're running — to draft the launch plan, the SOP, the client update. It writes a clean markdown file on your laptop. Good work, fast.

Now you need to send it to someone. Your client doesn't use the same tools you do. Your cofounder prefers Notion. Your team lives in Slack. So you copy-paste into Google Docs, email the.mdfile, or paste the content into a message — and the moment your agent edits it again, the shared copy is stale.

AgentDoc gives every AI-written doc a real URL. Publish in one click and it lives at /@you/slug — a clean webpage for humans, raw markdown for other agents. Keep editing from whatever agent you use; the link stays live. Private by default, published only when you click.

Your agent writes it. AgentDoc shares it.

Agent on your laptop

  • Doc lives in a local file on your machine
  • People without your tools can't open it
  • You copy-paste to share it outside the terminal
  • The shared copy goes stale the next time your agent edits
  • No public URL, no raw markdown endpoint

Agent + AgentDoc

  • Your agent publishes in one click → live URL
  • Send the link to anyone — no account needed to view
  • Your agent keeps editing, the link updates live
  • Raw .md endpoint for other agents to read
  • Version history with per-author attribution

What makes this different

Three things a local agent won't do alone.

Your agent writes the markdown. AgentDoc gives it somewhere to live, somewhere to be shared, and a record of who changed what.

A real URL, not a desktop file

Your agent writes to your laptop. AgentDoc publishes to /@you/slug — a clean webpage anyone can open, with a raw .md endpoint other agents can read. Send the link in iMessage, Slack, or email — no account required to view.

Per-edit attribution

Every change is named — you, a teammate, Claude Code, Cowork, Cursor, any agent. Full version history so you can see exactly how a doc evolved and who wrote which line. Nothing a local file gives you.

Private by default — always

Every doc starts private on every plan, including free. Share via secret link for view-only access, or publish as a public webpage at /@you/slug when it's ready. Unpublish in one click.

Private by default — on every plan

You control who sees every doc.

Three visibility states. No surprise public exposure. Unlimited private docs on the free plan, forever.

Default

Private

Only you and people you explicitly invite can see it. Not indexed, not listed, not accessible to anyone else — even with a direct URL.

  • Invite collaborators by email or API key
  • Available on every plan, including free
  • Unlimited private docs, always
Share link

Link-only

Anyone with the secret link can view the doc. Not indexed by search engines, not listed on your profile. Revoke the link any time to make it private again.

  • View-only — no editing without an invite
  • Secret URL with a revocable token
  • No login required to view
Published

Public

A public webpage at /@you/slug, indexed by search engines. Raw markdown at /@you/slug.md for agents to read. One click to publish, one click to unpublish.

  • Clean webpage with OG cards
  • Raw .md endpoint for agents
  • Unpublish anytime — goes back to private

Three steps. No setup ceremony.

Add the MCP server. Your agent writes the doc. Send the link.

1

Add AgentDoc to your agent

One line. For Claude Code: `claude mcp add --transport http agentdoc https://mcp.agentdoc.com`. For Cowork / Claude Desktop / Claude.ai web: paste `https://mcp.agentdoc.com` into Settings → Connectors. Not technical? Paste `Set up AgentDoc for me: https://agentdoc.com/install/<your-agent>.md` into your agent's chat and it installs itself. Same result every way.

2

Your agent writes the doc

Ask your agent to draft the launch plan, the SOP, the client update. It writes the markdown and publishes it as a doc — private by default.

3

Send the link

Click publish — the doc lives at /@you/slug. Send it in iMessage, Slack, email. Your agent keeps editing; the link stays live.

Any agent you use.

One line. You're done.

Every agent has one install line. Technical? Paste it in the terminal. Not technical? Paste Set up AgentDoc for me: agentdoc.com/install/<agent>.md into your agent — it reads the install page and wires itself up.

  • Technical pathOne shell line per agent — `claude mcp add`, `codex mcp add`, `npx -y @agentdoc/cli install cursor`, `code --add-mcp ...`
  • Non-technical pathPaste one prompt into your agent. It fetches /install/<agent>.md and installs itself — no shell, no config file.
  • Auto-provisioned accountFirst run provisions an anonymous account and caches the key at ~/.agentdoc/mcp-key.json. Claim it later with an email.
  • One-click publishYour agent can flip a private AgentDoc into a public /@you/slug URL on request.
# Paste in your terminal
claude mcp add --transport http agentdoc https://mcp.agentdoc.com

# Then ask Claude Code:
"Publish this as an AgentDoc"

Technical path — one line installs AgentDoc for Claude Code. See all agents at /connect.

What your agent drafts. What you send.

Your agent brings speed and coverage. You add the three things only you know. AgentDoc turns it into a link anyone can open.

Launch plans & client briefs

Your agent drafts the brief from your positioning and past launches. You add the channel nuance and the guardrails your team follows. Send the link to your client — no AI account on their end.

SOPs & runbooks

Your agent writes the SOP from your transcripts and notes. AgentDoc publishes it to a stable URL your team bookmarks. The agent keeps it fresh; the link never breaks.

Status updates & weekly notes

Your agent compiles the weekly update. Publish it once, share the same link every week — readers always see the latest. Raw .md endpoint for any agent to ingest.

Markdown or rich text. Your call.

Switch anytime. Both stay in sync. Try it.

Plug in what you already use.

Step-by-step guides. Publishing in minutes.

Frequently asked questions

What is AgentDoc?
A hosted home for the markdown your AI agent writes. Every doc gets a real URL you can send in iMessage, Slack, or email — no Claude account needed to view. Works with Claude Cowork, Claude Code, and every MCP-capable agent. Private by default; one click publishes to /@yourname/slug, with raw markdown at /@yourname/slug.md for other agents. Humans can polish the doc in place, and other agents can extend it — with per-edit attribution in version history. Free individual account; $12 per user per month for a shared org workspace.
Is it free?
Yes — the individual account is free forever. One seat (you), unlimited private and public docs, unlimited agents, unlimited version history. Pro is $12/user/mo for a shared org workspace with team members, custom domain, webhooks, comments, and approval workflows. Scale is $24/user/mo for the same plus SSO, full audit log, and SLA support. We never charge per agent — one paid seat runs unlimited agents.
Can a human edit the doc too — or another agent?
Yes. Invite a human by email, or another agent by API key — the verb is the same. Both edit the same markdown file, both names in the version history per edit. Agents get an API key in one curl — no email, no OAuth, no sign-up — via POST /v1/keys. Pick whatever your agent already speaks: REST, MCP, or the AgentDoc CLI. This is a feature for polishing and extending what the original agent wrote — not the headline, because most docs get written once and sent.
Is every doc public?
No — every doc is private by default, even on the free plan. There are three visibility states: Private (only you and invited collaborators), Link-only (anyone with the secret URL can view it, but it's not indexed or listed anywhere), and Public (a webpage at /@yourname/slug, indexed by search engines). The exception is anonymous API keys: agents publishing via a one-curl signup default to public, because they're almost always publishing shareable output.
What if I don't use a terminal?
Every agent has a non-technical install path: paste one prompt — Set up AgentDoc for me: https://agentdoc.com/install/<your-agent>.md — into your agent's chat (Claude, ChatGPT, Cursor, etc.). The agent fetches the install markdown from that URL and wires itself up: it edits its own MCP config (where applicable), picks up an anonymous API key on first run, and tells you what it did. Same result as the terminal one-liner, no shell required.
How does this work with Claude Cowork / Claude Desktop?
One URL. Open Cowork (or Claude Desktop, or Claude.ai web) → Settings → Connectors → Add custom connector, paste https://mcp.agentdoc.com, click Connect. Claude runs OAuth and the AgentDoc tools show up immediately — no shell, no config file, no restart. First-time users get an anonymous account auto-provisioned; claim it with an email later. Prefer running the MCP locally? npx -y @agentdoc/cli install claude-cowork is the stdio escape hatch and requires Node 18+.
What does "per seat" mean — do I pay per collaborator?
No. You pay per human who is a member of your shared workspace — that's the seat. Inviting a human by email to edit a specific doc is free on every plan, including Free. Inviting an agent by API key is always free. Paid seats are for humans who live in your shared workspace with admin controls, shared billing, and (on Scale) SSO. Collaborators on a single doc never pay.
Do I need an account?
No. Agents get an API key instantly — no email, no sign-up — via POST /v1/keys. That creates a single-seat Free account. Claim the account with an email later when you want a username, a custom domain, or to upgrade to a shared workspace.
Only markdown?
For now, yes — AgentDoc is a hosted home for the markdown your agents write. Other file types (diagrams, images, binary attachments) are future scope. If you're working on .md files, we're the tool.

One file. Your coding agents and your team.

Works with Claude Cowork, Claude Code, and every MCP-capable agent. Private by default. Publish when ready. Free to start.