please run

Figures out how to start this project, then remembers it.

please run
please run set [--term] [--check "<probe>"] "<command>" ["<command>" ...]
please run reset

The first time, please run reads a small, fixed set of files if present, README, package.json, Cargo.toml, pyproject.toml, requirements.txt, Dockerfile, docker-compose.yml, Makefile, go.mod, Procfile, and .tool-versions, and asks the AI how to start the project. It shows you the plan and asks for confirmation before running anything.

Once confirmed, the decision is saved outside the working tree, so it never shows up in git status, and every please run after that replays it directly with no AI call and no confirmation.

If the project needs something running first, such as a database or a Docker daemon, the plan can include a cheap check that runs before the real commands, with a short hint printed if it fails.

Dev servers open in a new terminal#

If the run command is a long-running process, a dev server or a file watcher, please run opens it in a new terminal window so your current shell stays free. Pass --here to force it inline, or --new-terminal to force a new window.

please run set#

please run set "npm run dev"
please run set --term "bin/dev"
please run set --check "docker info" "docker compose up -d" "npm run dev"

Records the run commands by hand, with no AI call. This is also how please chat overrides how a project runs: tell the agent what your project actually needs and it saves it with please run set. --term marks the command long-running; --check gives a probe command that must succeed before the rest.

please run reset#

please run reset

Clears the saved decision, so the next please run figures it out fresh.