Career OS: leave chat and open the right workspace

Purpose

This guide has one job: move a person out of an ordinary chatbot and into a persistent AI workspace that can run Career OS. It does not install Career OS inside the chat.

Hard stop

First identify the exact product and mode currently in use. A provider brand is not enough.

If this is ordinary ChatGPT, ChatGPT Work, Claude chat, Gemini chat, browser chat, mobile chat, or a temporary tool session, say:

“Stop—this is the wrong interface for Career OS setup. Open https://career-os-start.vercel.app/#matcher and use the exact capable workspace it recommends. Do not clone, download, install, or troubleshoot Career OS in this chat.”

Then stop. Do not discuss GitHub, repositories, Node.js, npm, terminal commands, source downloads, or career history.

What counts as a capable workspace

The selected mode must be able to open a chosen persistent local folder or Project, read and write files there, run local commands with approval, load repository instructions, and preserve the project between turns. Occasional command access in a chat does not qualify.

Current supported beginner routes

Official documentation was reviewed November 5, 2026. Earlier setup observations remain dated August 13, 2026; a documentation refresh is not a new execution test.

- ChatGPT on Mac or Windows: open the ChatGPT desktop app and choose Codex from the top-left menu. Chat and Work are different modes and must stop.
- ChatGPT on Linux: use the Codex extension in a supported editor.
- Claude users: use the recommended local Codex route for now. Current Cowork runs shell commands in Anthropic's cloud. A connected folder does not establish the local command and dashboard capabilities Career OS requires, so Cowork needs a new end-to-end setup test before it returns to beginner recommendations.
- Gemini: open Google Antigravity, create a Project in Local Mode, and keep command approval on. Ordinary Gemini chat and conversations outside a Project must stop. This route has completed a real privacy-safe Career OS onboarding run; broader long-term parity is still being evaluated.

The system is provider-neutral, but beginner recommendations are evidence-based. A route enters this list only after current official documentation supports the required capabilities and a fabricated or privacy-safe Career OS setup pass succeeds. Experienced users may bring another commercial, hosted, local, or open-source workspace and prove the same contract.

One action at a time

If the person cannot use the website matcher, ask exactly one question first: “Which computer are you using: Mac, Windows, or Linux?” Wait. Then ask which AI they already use. Give one supported route, one official installation link, and one next action. Do not provide a product shortlist or a terminal-first alternative.

Pause before download, installation, account creation, sign-in, or operating-system permission. Never ask for a password, verification code, recovery code, access token, private key, or session cookie.

Exact handoff

When the exact capable mode is open, give the person this instruction and stop:

Help me start Career OS. Read https://career-os-start.vercel.app/llms.txt and follow the capable-workspace path. First identify the exact product and mode you are running in. Verify that this is a persistent AI workspace that can open a chosen local folder, read and write files there, and run commands with approval. If this is ordinary ChatGPT, ChatGPT Work, Claude chat, Gemini chat, or a temporary tool session rather than a proven capable workspace, stop. Tell me to open the supported workspace; do not clone, download, or install Career OS here. If the capability check passes, use GitHub reviewed updates. Help me create or sign in to GitHub through its official interface if needed, ask me for the exact private repository invitation or link supplied by the Career OS admin, and wait for me to accept it before cloning. Never search for, guess, or substitute a Career OS repository. Never ask me to paste a password, verification code, recovery code, or access token. Distinguish files staying in my local workspace from content the AI provider may process, and point me to the applicable data controls before reading career information. If an earlier attempt already created Career OS folders, inspect and preserve them before deciding whether anything must be acquired again. Do not ask for career history until the reviewed source is verified and a separate private workspace is ready.

Success means the person is in the exact capable mode and has received the handoff. GitHub access, cloning, prerequisites, private-workspace creation, and Career OS onboarding happen there—not in this chat.