Skip to content

PlanExe2

PlanExe2 turns a planning idea into a strategic plan draft: a single HTML report that leads with the decisions you still have to make, backed by the supporting analysis and an audit trail of what was checked.

PlanExe v1: generate an unusually comprehensive expert planning workshop.

PlanExe2: construct a partially validated model of the project, then expose what humans still have to decide.

PlanExe2 is PlanExe v1's plan-generation pipeline, rewritten as skills executed by a small, dependency-free Python DAG runner.

  • No pip install. Python 3.11 or newer, standard library only.
  • LLM calls go through the headless Claude Code CLI (claude -p), on your Claude subscription.
  • You don't write the prompt yourself: your coding agent interviews you, drafts the prompt, checks it and launches only after you confirm.
  • Every stage writes plain JSON and markdown files. Edit any of them and re-run: your edit is kept and everything downstream is regenerated.

The output is a draft to refine, not ground truth. It surfaces the questions and decisions a plan depends on.

Pages

  • Getting started: requirements and how to make your first plan.
  • The report: what the report contains and how to read it.
  • Commands: running PlanExe2 by hand, options, stopping and resuming.
  • How it works: the DAG, skills, and how edits are picked up.

PlanExe v1

The v1 documentation (MCP server, web frontends, Docker, OpenRouter/Ollama/LM Studio providers) is in the PlanExe repository. PlanExe2 does not have these components.

Source: github.com/PlanExeOrg/PlanExe2.