Personal AI · free to try

Your personal AI

An AI workspace that learns about your world through readable files, corrections and everyday work.

Kasv.ai About 2 min read Version 2026-10-25

Download a clean starting point

Personal AI is a file-based AI workspace. The package contains readable Markdown instructions, empty memory structures and optional workflows. It runs with a file-capable AI agent such as Codex; it does not include a language model or a standalone chat application.

Clean starter package · beta Download Personal AI ↓ ZIP · 105 files · includes ALUSTA-SIIT.html · Estonian-language base

Start with Codex

  1. Extract the ZIP into a new, separate folder.
  2. Add the extracted Personaalne AI folder as a local project in Codex and start a new chat in that project.

Memory, privacy and your control

By default, the workspace saves conversations and useful knowledge into local files. You can inspect those files and ask the agent to correct or forget information.

What changes with a personal workspace?

  • Less repeated background. Keep useful project context and earlier decisions in files that the agent can read.
  • Correctable memory. Inspect and change the information the workspace keeps about you.
  • Reusable ways of working. Keep working rules and lessons from previous attempts across separate tasks.
  • Explicit agreements. Tell the agent what it may do, where it must ask and when to stop.

What is inside?

The AI folder holds working instructions. The registrid folder starts the editable knowledge base; vestlus holds conversation logs; projektid holds real work. Optional paths cover imports, reflection and questionnaires. The README explains the starting point. Messenger import instructions are included in the package: ask your AI to help import your Messenger conversations, and it will read the local instructions and guide you through the process.

I develop and use my own workspace, Sofia, with Codex. Personal AI is the clean starting framework that others can inspect and adapt. It carries the working structure, without inheriting my personal history.

The package is an experimental starting point. Try one small task, check what changed in the files and decide whether the approach is useful for you.