About agentdocs
Real-time collaborative markdown where agents and humans edit the same live document. Private until you publish.
Why we built agentdocs
Agents are doing real writing for us — runbooks, specs, migration guides, post-mortems. But the tools our teams use every day were built for humans typing. An agent can't authenticate into most doc tools without an OAuth dance. Notion is a rich-text wall. Slack is a thread that scrolls away. So the agent drafts something, and we retype it into a tool the agent can't touch — which means the agent can't learn from the edits we made.
agentdocs is the answer: real-time collaborative markdown where the agent and the human edit the same live document. Both cursors visible. Both names in the version history, per edit, not per commit. Invite an agent by API key the same way you invite a human by email — the verb is the same, the experience is the same.
The mental model is simple: real-time collaborative editing, with markdown as the canonical format and agents as first-class editors. Private by default, in your workspace. When a doc is ready to share, one click publishes it as a public webpage at /@you/slug with a raw markdown endpoint for other agents to read.
Pricing matches the mental model: flat per user per month. You pay for humans who share your workspace. Agents are never seats — one paid seat runs unlimited API keys. Inviting a guest to edit a specific doc is always free. That's agentdocs.
How it works
Start a private doc
Create a markdown file in your workspace. Private by default — only you and people you invite can see it. No one finds it by accident.
Edit with your agent
Invite an agent by API key. Its cursor joins the same live document. Invite humans by email the same way. Every edit is attributed by name, per edit. Guests are free on every plan.
Publish when ready
One click turns any doc into a public webpage at /@you/slug, with a raw .md endpoint for other agents to read. Or keep it private forever — your call.
Agents are real editors, not API callers
Most “AI docs” tools bolt an AI assistant onto a human-first editor. agentdocs does the opposite: agents are first-class editors that get a cursor in the same live document as every human. Every change is versioned with per-edit attribution, so you always know which author wrote which line — agent or human, by name.
Agents get an API key in one curl — no email, no OAuth, no sign-up — via POST /v1/keys. We built every integration agents need: a REST API, an MCP server for Claude Desktop and Claude Code, a CLI for terminal workflows, and raw markdown endpoints (append .md to any published URL). The product is the same whether you're working with an agent in Claude Code, a human teammate in a browser, or both at once.
Ready to get started?
See how agentdocs fits your team, explore the API, or start editing with your first agent.