Switch between Kimi Code accounts

The short answer

yoink add
yoink save kimi-personal --tool kimi
yoink use kimi-personal
yoink use kimi-personal --force
yoink current --tool kimi

yoink add handles the sign-in, yoink save captures the account Kimi Code is signed in to now, and yoink use switches. Add --force when you switch from a script while Kimi Code is still running. The output names the account by its Kimi user id:

✔ Switched to kimi-personal (kimi:u1001)
Restart Kimi Code to pick up the new login.

Why kimi login cannot add a second account

kimi login does not offer to add another account while one is already signed in. yoink works around this by keeping a copy of the current login as a profile, then giving kimi login an empty folder to sign in to.

Kimi Code keeps its login home at $KIMI_CODE_HOME when that variable is set, otherwise at ~/.kimi-code. The login is the set of JSON files in its credentials/ folder. yoink snapshots every one of them, so a profile is the whole folder's JSON content rather than a single token.

To name the account, yoink decodes the access_token in those files locally and reads its user_id claim, checking credentials/kimi-code.json before the others. The label is kimi:<user_id>, and if the token carries a region claim, yoink shows it next to the id. If no file holds a readable user_id, yoink stops with "Could not identify the Kimi Code login" and asks you to run kimi login again, instead of saving a profile it could not tell apart from the others.

Add an account with yoink

Run yoink add in a terminal and choose Kimi Code. Here is what happens to your files:

1. yoink saves the live login as a profile, if no profile holds that account yet.
2. It renames credentials/ to credentials.yoink-<timestamp> in the same Kimi Code home. Nothing is deleted.
3. It runs kimi login, which now sees no account and starts a fresh sign-in.
4. When the sign-in succeeds, yoink captures the new login and asks for a name. Any non-JSON entries from the old folder are moved into the new credentials/ (unless one of the same name already exists), and the set-aside folder is removed.

If kimi login fails or produces no login, yoink deletes whatever partial credentials/ it left and renames the set-aside folder back, so you end up signed in as before. yoink also tells you which profile still holds the previous login, should restoring it ever fail. yoink add needs an interactive terminal and the kimi command on your PATH.

Switch

yoink use kimi-work re-saves the live login into the active Kimi Code profile (only when its user id matches), writes the target's credential files with owner-only permissions, and removes any live credentials/*.json file the target does not have. Kimi Code alone changes; Claude Code and the other tools keep their logins. If a kimi process is running, yoink asks first in a terminal and refuses in a script unless you pass --force. The shared rules are in What a switch does.

Kimi Code login vs the Kimi Code API key preset

Two different things carry the Kimi Code name in yoink, and they do not overlap:

yoink never copies a subscription token into a harness config. To use Kimi models in Claude Code, opencode or another agent, add the preset as a provider; the preset table lists its endpoints, and the harness table shows which protocol each agent uses.

FAQ