Run a second Claude Pro or Max subscription in Claude Code with the same CLAUDE.md, skills, memory, hooks, settings and chat history. Only the login changes. No proxy: each account signs in through Anthropic's own page.
Copy this prompt and paste it into Claude Code. Claude downloads the skill, checks its fingerprint, reads the code and checks what it claims, then runs the setup. The one thing you do yourself: sign in to your second account in the browser.
claude-subs.zip sha256 c328def1b9d7ce2b089b0032e17285f4a85f4b0134830c58ce13afaf86d6b8b7
You need a second Pro or Max plan on a different email. When the browser opens for the sign-in, check the email on the page: it must be the second account. If it shows your main one, switch account there or use a private window.
This version carries both files inside the prompt (45 KB) and checks their fingerprints before running anything. It works offline, but Claude has to write about 44 KB of code out by hand, which took about 9 minutes in our test.
| Type this | What happens |
|---|---|
claude or claude-a | Your main account, same as today. |
claude-b | Your second account, same brain. |
claude-auto | Starts whichever account has the most room left in its 5-hour and weekly windows. |
/exit then claude-b -c | A limit hit mid-task. The same chat keeps going on the other account. |
claude-sub status | Both accounts: email, last-seen 5-hour and weekly use, and which one it would pick. |
claude-sub exec b -p "..." | Scripts and cron on account b. It bills the plan, never an exported API key. |
Claude Code reads everything from one config folder and keeps that folder's login in its own slot. So a second folder is a second account. The second folder links back to the first for everything except the login.
Arrows are links into ~/.claude, so a change made from either account lands in the same file. On macOS each login lives in its own keychain item, Claude Code-credentials-<8 hex of the folder path>. On Linux it is the folder's own .credentials.json. The tool never reads either.
A proxy like CLIProxyAPI sits between Claude Code and Anthropic, holds the login tokens of several accounts, and spreads requests across them. Anthropic's Claude Code terms rule that out:
"...developers may not collect, store, or intermediate Claude.ai credentials or session tokens"
"sign-in to a Claude account must complete through Anthropic's own flow."
The same page allows "an end user signing in to the unmodified Claude Code binary with their own Claude subscription". That is all this does. Proxy users have reported accounts shut off with "This organization has been disabled". Rules change, so read the current page: code.claude.com/docs/en/legal-and-compliance.
/mcp in a session on that account and sign in once.ANTHROPIC_API_KEY exported, a plain claude -p bills that key. claude-a -p, claude-b -p and claude-sub exec drop it for the run so the plan pays.It signs account b out, moves ~/.claude-b aside, and when b was the last extra account, removes the shell block and restores your old status line. Nothing is hard-deleted.