04 / Clean-machine setup

New VM.
Two secrets.
Go.

You provide the proxy base URL and API key once. Setup discovers the live model registry, installs Pi and subagents, writes secure local configuration, chooses model roles, and verifies the connection.

Install from this folder

$ cd /path/to/loop-maxxing
$ ./scripts/install.sh

LLM API base URL: https://your-proxy.example
LLM API key (hidden): ••••••••••••

Install after publishing

$ npm install -g @sudosapient/product-loop
$ product-loop setup
Automation / CI: set LLM_PROXY_BASE_URL and LLM_PROXY_API_KEY in the VM’s secret store, then run product-loop setup. The CLI intentionally has no --api-key flag, so credentials do not leak into shell history or process listings.

Start your first project

$ mkdir tiny-product && cd tiny-product
$ product-loop init
$ $EDITOR PRD.md
$ git add PRD.md && git commit -m "Define product outcome"
$ product-loop start

What setup changes

LocationAction
~/.config/product-loop/envStores URL + key with owner-only permissions and sources it from your shell profile.
~/.pi/agent/models.jsonAdds or updates only the llm-proxy provider; preserves other providers and creates a backup.
~/.pi/agent/settings.jsonSets the proxy provider and a default model only when a default model is absent.
~/.config/product-loop/model-route.jsonRecords selected parent, worker, and reviewer IDs—never the API key.