Switch between Claude Code accounts

The short answer

yoink add
yoink list
yoink work
yoink use personal
yoink current

yoink work and yoink use work do the same thing: any argument yoink does not recognize is treated as a profile name. yoink current prints the active profile for each tool (add --tool claude for Claude Code alone), and yoink list shows every saved profile with the active one marked.

What identifies a Claude Code login

A Claude Code login has two halves, and a switch has to move both of them:

A yoink profile stores the pair, so restoring it gives Claude Code a complete login rather than a token that disagrees with the account shown in ~/.claude.json. Profiles live in ~/.config/yoink/profiles.json, which yoink keeps owner-only (chmod 600). How it works has the full picture.

Save each account once

Run yoink add in an interactive terminal and pick the Claude account option. yoink first saves the login that is live right now (if no profile holds it yet), then runs claude auth login --claudeai so you can sign in to the next account in the browser. When the sign-in finishes, yoink captures the new login, suggests a profile name based on the account email, and asks whether to add another.

A common setup is work for your company plan and personal for your own. If the login did not change after the sign-in, yoink says so and saves nothing, so you cannot end up with two profiles for one account by accident.

Already signed in and only want a snapshot of the current account? yoink save temp captures the live Claude Code login under the name temp.

Switch from the terminal or the menu bar

From the terminal, yoink <name> switches directly, and plain yoink opens the interactive menu, where Enter switches to the highlighted account. On a Mac, the menu bar app lists the same profiles, because it runs its own copy of the CLI against the same profiles.json.

On macOS the first switch triggers a Keychain permission prompt. Choose Always Allow so yoink is not asked again on every switch. The write itself goes through security -i, with the command sent on stdin and the credential hex-encoded, so the token never shows up in the arguments of a running process.

Why the active login is re-saved first

Claude Code refreshes its token in the background, so the copy in your profile can go stale while you work. Before every switch, yoink reads the live credential back out and stores it in the active profile. It only does this when the live oauthAccount email (and account id, when both sides have one) matches that profile, so a login that belongs to another account is never written into the wrong profile. If writing the new identity fails halfway, yoink puts the previous live login back.

Switching while Claude Code is running

A running session can write its old token back on its next refresh, which would undo the switch. When Claude Code is running, yoink use asks before switching, with "no" as the default. In a script with no terminal attached, it prints a warning and switches anyway, so quit Claude Code first and start it again after the switch.

Doing it by hand with CLAUDE_CONFIG_DIR

Some people skip a switcher and give each account its own config directory, setting CLAUDE_CONFIG_DIR per shell. yoink does not manage those directories for you, but it respects the variable: when CLAUDE_CONFIG_DIR is set, yoink reads and writes .claude.json and settings.json inside it, matching Claude Code. On Linux and Windows it also reads and writes .credentials.json there. On macOS the credential stays in the single Claude Code-credentials Keychain entry whatever the variable says, so per-directory setups on a Mac still share one stored login. The comparison of Claude Code account switchers covers the other tools that take each approach.

FAQ