The notes before the prompt.

Give Claude the page, not a paste

Connect Claude once and it reads your pages itself, as markdown, heading by heading. Ask about “the risks in the checkout plan” and it opens that page; the plan stays the one place it lives, instead of a copy in a chat that goes stale the moment someone edits the original.

From a spec to a plan

Start from the Plan template: the goal, milestones as a checklist, the risks, and what done looks like. Ask Claude to draft the milestones from the spec beside it and it rewrites that section alone, while anyone else on the page keeps typing. Its words land live with a short glow on what it added, and History lists them in Activity beside people’s edits.

One Claude, wherever you use it

Claude Code, Claude Desktop and claude.ai each connect with the same sign-in and work as one agent, Claude, with its mark. People on the page see it among them while it writes. When two people’s Claude have been on a page, each shows its person: Claude (Jane) beside Claude (Sam).

Ask it in the margin

Select a line and choose Add comment, then type @ and pick Claude. It finds the question the next time it looks, answers in the thread and fixes the line, and it can open threads of its own beside the text it means.

Connect Claude

Claude or another AI assistant you connect can read and add to your pages, as you: only the pages you can open, and never more than you could change. It can’t delete for good, share or publish. On claude.ai, add Antescript as a connector; in Claude Code, one command adds it. Free includes 100 assistant calls a day, and Agents has the steps.