What is plan-tree?

Plan Tree is a portable planning tool designed to help AI agents and developers keep track of project decisions, open questions, and progress across many sessions. Most AI tools have a built-in planning feature, but those plans are short-lived: they tell you what to do next but forget why certain paths were chosen, what was rejected, and where a project left off last time. Plan Tree tries to fix that by keeping all of that context in a structured set of Markdown files that can be read and updated by any agent or any person. The core concept is a folder of plain text files organized into stable categories: a roadmap, current implementation status, open questions, past decisions, topic notes, and a history archive. New information almost always fits into one of those categories without forcing a restructure. The README contrasts this with code, which has to carry both intent and executable behavior at once, making it harder to keep tidy as a project grows. The main way to use the project is to copy a short usage rule written in plain English into your agent's memory file or team configuration. Once that rule is in place, the AI assistant is supposed to consult the planning tree before starting any new work, record decisions and discoveries back into it after finishing, and stay in a clarification mode until the plan is concrete enough to act on without guessing while coding. The project includes a SKILL.md file (the actual rule definition), an OpenAI agents YAML configuration, and two reference guides covering maintenance patterns and migrating older planning documents into the new structure. A Chinese translation of the README is also provided. This is a small, opinionated workflow tool aimed at people who use AI assistants for software projects and find that context keeps getting lost between sessions. It does not generate plans automatically, it defines a place to store them and a discipline for keeping them current.