Plainpaper

Plainpaper is an infinite canvas for marketing campaigns, shared between a marketer and their AI agent.

If you have used Miro or Milanote, the surface is familiar: an infinite board, cards you drag, group and comment on, phases you can lay a campaign out across. The difference is who fills it in. Plainpaper is AI-agent first. Connect Claude, ChatGPT, Cursor or any MCP-compatible client, and the agent reads the board's current state, produces or revises work, and writes it back as typed cards: research, segments, insights, plans, emails, ad creative and results.

The agent arrives briefed for higher quality output without being the typical AI slop. A board is not an empty document. It carries the campaign process, the phases and the field structure of each artifact, so the agent knows what a finished marketing email looks like before it starts one. Above that sits a guidelines library holding your brand, voice and do-and-don't rules, which the agent must read before authoring anything. One example built-in block is a "Prevent AI language" preset with toggles for AI punctuation, hype, filler, formulaic structure and emoji, passed to the agent verbatim. Generic, off-brand output is a briefing failure, and every layer here removes one.

You see everything, and you hold the gate. Work lands as typed cards on a shared canvas rather than scrolling past in a chat window. You watch the campaign assemble, edit cards, and leave comments the agent picks up next session. At the moment you approve, the drawer shows the resolved lineage behind the card: the artifact, the segment it targets, the insight that informed it. Reversible work the agent does directly. Anything that spends money or cannot be undone is human-only, always, classified in server code and failing closed, so an unrecognised action can never slip through.

The reason it holds up on a real campaign is memory. Because the campaign state lives on the board and not in a chat transcript, the agent can open a clean context session and reconstruct the whole thing in a single call. Your agent's memory does not scroll away.

Two deliberate boundaries: Plainpaper is not an LLM, so there is no model lock-in and no token resale, and it is not the execution layer, so it never holds your email or ad platform credentials and can never spend your money.