The visual canvas
where the AI agent
runs your marketing

Plainpaper is the shared, persistent workspace for AI-driven marketing. It's the canvas where your AI agent builds a campaign (research, audiences, insights, plans, emails, creative, results) as structured, typed cards on one infinite board. The agent does the heavy lifting while you steer, edit, and approve. It lives outside the chat, so it stays readable and never scrolls away.

No. Plainpaper is not an AI provider. You bring your own agent via MCP: Claude Desktop, Claude Code, Cowork, ChatGPT, or any other MCP-compatible client. The model leads the work, and Plainpaper is the durable workspace those chats write into.

A board is both a working desktop and a proven playbook. Start one from a template (the Ecommerce Campaign board is live now) and the phases (Research → Audience → Strategy → Create → Launch → Measure), starter cards, and agent guidance are already in place. Your agent follows the methodology step by step while you watch it fill in. More official playbooks are on the way, and you can save your own as a private template.

Yes. Social ads are a first-class card type in four formats: single image, video, carousel, and vertical video. You review copy and creative in the shape they'll actually run, with placement previews, and approve them with the audience and insight that shaped them right there on the board. Images and video live alongside HTML email campaigns, drafts and production-ready assets side by side.

It doesn't, directly. Your tools connect to the agent through their own native MCP servers or APIs. Plainpaper is the context and memory layer, not the integration layer, and it never holds your credentials. When your agent pushes an approved email or ad, it calls that platform itself, and Plainpaper is where the asset and the rationale live.

On the board. Your agent writes results and figures back as Result cards, linked to the plan and assets they measured. The board doubles as a flexible dashboard during and after a campaign, and those numbers become the research that fuels your next board, on the same canvas or a fresh one.

Not if you don't want it to. Your workspace has a guidelines library the agent must apply before writing anything: tone of voice, brand colors, logo files, do/don't rules, and a "Prevent AI language" preset with toggles that cut the em-dash-and-hype tells. Set it once and every board that opts in gets copy that reads like you.

Your work is structured data, not a proprietary blob: card bodies are plain markdown or HTML, and connections are typed links. Your agent can read every card and asset out over MCP today, so nothing is locked in. A one-click workspace export is on the roadmap. No goodbye tax, ever.

Yes. Every account starts with a 7-day free trial: every feature, Starter's limits, no credit card. That's enough to build one real campaign end to end. If you decide not to subscribe when it ends, you move onto a free tier rather than losing access: one active board, with your agent still connected, enough to keep working on something while you make up your mind.

For the canvas itself, no. It's drag, click, and type. To connect your agent via MCP, you do a one-time setup in Claude Desktop, Cowork, or your client of choice that's about as involved as adding a browser extension. We'll walk you through it.