pal code

Input savers

Rules file (one pass) Step methods

Pasted text, code, URLs and attached files are never edited. A zip of source code is indexed as a graph; big documents are indexed too; only relevant excerpts are sent.

Local LLM: checking…

API key

Kept in this browser tab only. Or set the provider's env var.

Stats

Loading…

Memory

Checking…

Your messages are kept on this machine so later sessions can recall them. Secrets are skipped.

Studio

Saved this chat: 0 tokens (est.)0 API calls avoidedAPI used: 0 in / 0 outTokens left: …

Demo: plan a launch in 6 messages

Sends the same 6 messages twice to the model chosen in the header: once the way a plain chat app does it (whole history every time, no savers), once through Pal. Then compares the tokens the API reports and the time taken. About 12 API calls. It uses no memory and is not added to Stats.

Test the input savers on your own text

also call the API model, raw vs saved