Mapfable addresses the drift that appears when a project is built across many AI conversations: the scope changed somewhere, something is marked done but was never reviewed, and the next step is buried in a thread. It keeps the plan as a map of nodes with explicit status, so define-scope, build-core, review-login-flow and prepare-to-launch read as one track instead of as chat history.
Each node carries the decision and the evidence with it. The example project on the homepage shows entries such as a recorded boundary (build the personal version first, teams can wait, stop after the login review), an evidence note listing checks to run (sign in, sign out, retry after cancellation), and links to the working files the plan grew out of, like project-notes.md and build-log.md. It also separates done from verified, asking whether something is finished or still awaiting review before launch.
The map is meant to be read by your AI as well as by you: the site describes connecting your agent to the saved context so a build continues from where the map stands, rather than from a re-paste of the previous conversation. The interactive example on the homepage is explicitly labeled so visitors know their own projects are not changed by trying it.
Access is through Google sign-in; flowcharts are listed as a separate mode and a Chinese version of the site is available. The homepage publishes two plans, $9 a month or $85 a year in US dollars, and states that once trials are available the first Google sign-in starts a 7-day trial with no card and no automatic charge.