Skip to content

Getting started

Requirements

  • Python 3.11 or newer. Standard library only; nothing to pip install.
  • The Claude Code CLI, logged in: claude auth login. PlanExe2 makes its LLM calls through it, on your Claude subscription.
  • Usage budget: a full plan is about 220-260 LLM calls and takes roughly 65-90 minutes.
git clone https://github.com/PlanExeOrg/PlanExe2.git
cd PlanExe2
claude

Then ask for a plan, for example "make a plan for a bakery in Lyon". Codex and other agents follow the same steps via AGENTS.md.

What the agent does

The make-plan skill (.claude/skills/make-plan/) takes over:

  1. Asks a few questions about what the plan needs and you haven't said yet: objective, location, budget, timeline, scale, stakeholders, constraints, success criteria. Each question comes with suggested answers to pick from, and "you decide" where a sensible default exists. It also asks for the plan's start date (Month 0); past and future dates are fine.
  2. Drafts the prompt: 300-800 words of prose built from your answers. Anything it assumed on your behalf is labelled as an assumption, so the plan treats it as an estimate.
  3. Checks the prompt with check-prompt (one LLM call). If important items are missing, it asks again.
  4. Asks for confirmation. It shows the prompt, the check result, the start date and the cost, and launches only after you say yes.
  5. Runs the plan in the background and reports back when it is done, starting with the decisions you need to make.

While it runs

Progress lines show the current stage and an ETA based on the DAG's critical path. The same data is in runs/<yyyymmdd>_<name>/.planexe_skill/progress.json. If something fails, re-running resumes where it stopped; finished stages and completed LLM calls are kept.

The result

Plans are written to runs/<yyyymmdd>_<name>/ inside the clone (ignored by git), for example runs/20261008_cross_border_rail_ticketing/. Open report.html; report.md has the same content. See The report for what is in it.

Update with git pull. Each report's Metadata section names the PlanExe2 version that produced it.

Without git: let the agent fetch it

In any Claude Code session, paste:

Clone https://github.com/PlanExeOrg/PlanExe2, read its CLAUDE.md, and use its make-plan skill to make me a plan for a bakery in Lyon.

This works, but a session started outside the repo does not load the repo's skill on its own, so the agent has to find and read CLAUDE.md first. Cloning and starting the agent inside the repo is more reliable.

Sandboxed agents

The child claude process needs keychain access for authentication. If your agent runs commands in a sandbox (for example Claude Code desktop with sandbox on), the plan run must happen outside it. If authentication fails, run claude auth login and resume.

Why there is no pip package

There are no dependencies to install, and the skills, prompts and schemas live in the repo next to the agent instructions that drive them. A pip package would move the make-plan skill out of your project and need a separate install step for the guided flow. A Claude Code plugin is the likely route if PlanExe2 should work from any folder without cloning.

Next