Resume a campaign

The board is the memory

Every chat has an ending: the transcript scrolls away, the session expires, the context window fills up. A campaign that lived in the chat dies with it. A Plainpaper campaign survives, because everything your agent made, what state each piece is in, and how the pieces connect all live on the board rather than in the conversation.

The test Plainpaper is built to: a clean session, with no chat history at all, must be able to reconstruct the full state of a campaign from the board, down to every artifact, its lineage, and its status.

That's what "resume" means here. A brand-new session reads the board and knows the campaign, without scrolling up or pasting old messages back in.

It knows more than a transcript ever held, in fact. A chat remembers what was said; the board holds what was made: each card with its status, and the connections that say this email targets that audience and follows that one in the send order. That structure is what the next session stands on.

The resume prompt

Open a fresh session with your agent (tomorrow, next week, after two weeks of holiday) and ask:

Open my board "Summer sale" in Plainpaper and tell me the state of this campaign: what's drafted, what's approved, what's still open, and what you'd do next.

The agent reads the whole board in one step (the phases, every card's one-line summary and status, and the connections between them) and narrates it back: what's drafted, what's approved, what's still open, and how the artifacts relate. That's your standup, delivered by the agent, from the board.

Try it deliberately once, early on: end a session in the middle of real work, open a fresh one, and ask. When a brand-new session describes your campaign accurately, with no chat history at all, the board stops feeling like a place you file things and starts feeling like the memory.

Continue the work

Because the sequence and the lineage are on the board, the next instruction can be almost lazy. The emails in a sequence are chained in send order, so "draft the next email" isn't a guess. The agent follows the chain to see what came before, applies your workspace guidelines, and writes the next proper email card into the right phase.

Draft the next email in the sequence on my "Summer sale" board. Read the board first: follow the send order, stay consistent with the emails that came before, and apply our workspace guidelines. Save it as a draft for my review.

"Proper" is doing real work in that sentence: what comes back is a structured email card (subject, preheader, rendered body) in the right phase, saved as a draft for your review pass. See Deliverables for what makes an email card an email card.

Keep the summary honest

The board summary is the first thing a fresh session reads, so it pays rent every time you resume. Make updating it the last move of every working session:

Update the board summary for "Summer sale" with where we stand and what's next, in plain language. Then check the card summaries: if any no longer match their card, fix those too.

Card summaries matter the same way. Every card carries a one-line summary, and those lines are what let the agent read a whole campaign cheaply instead of pulling every card body. Plainpaper even tracks when a card's content has changed after its summary was written, so a stale summary is visible rather than silently misleading the next session.

Anywhere, any client

The board is the source of truth rather than any one chat window. So you can brief your agent from Claude Desktop at your desk in the morning and follow up on claude.ai from the sofa that evening: the same board and the same memory, with no handover. The same goes for any connected MCP client: whichever agent opens the board next reads the same state as the one that wrote it.