v1.0.0 · free and open source
Handova runs your AI coding agent under as many accounts as you have. When one runs out, it hands the conversation to another in one command.
npm install -g handovaClick a command to run it.
AI coding agents like Claude Code hold one login at a time. Switching means logging out and back in, and when an account runs out mid-task, the conversation and everything the tool learned about your project stay behind with it.
Handova gives every account its own login, so they all stay signed in, and moves a conversation to whichever one you need to keep going.
What it does
Each account gets its own folder and login for your coding agent. Logging in or out of one never touches another, and you can run them side by side in different terminals.
handova workWhen one account hits its limit, Handova copies the conversation and that folder's memory to any account you choose, then opens it there so you carry on.
handova handoff personal workMemory that already exists on the receiving account is kept. A conversation that was continued on both copies is left alone, and Handova tells you where to resume it.
handova handoff personal work --pickNew Claude Code accounts share your CLAUDE.md, skills, agents and commands, so every account works the way you set it up. Settings are copied, so each account can have its own.
handova add workHandoff
Handing a conversation to another account means it's processed under that account's organization. Handova names the organization and asks before it copies. If your workplace doesn't allow moving work between accounts, don't.
Commands
Opens the account's coding agent, today Claude Code, logged in as work. Anything after the account name works the same as it does with plain claude.
Run it in the folder you were working in. It finds your latest conversation there on personal, names the organization it's going to and asks first, copies the conversation and memory to work, and opens it there. Any two accounts work: handova handoff work client, or back again with handova handoff work personal.
Creates ~/.claude-client, shares your CLAUDE.md, skills, agents and commands with it, copies settings.json, then opens the login.
Shows every account with its email, organization and folder.
Forgets the account and keeps its folder and login. With --purge, it also logs the account out and deletes its folder, after asking.
Checks that each account's coding agent is installed, every account is logged in, and the shared links still work. For anything wrong, it says how to fix it.
Prints a claude-<account> shortcut for each account. Add eval "$(handova shell-init)" to your shell profile to use them.
No. Each account keeps its login in its own place, so they all stay signed in and you can run them at the same time in different terminals.
Not yet. Handova supports Claude Code today and is built so other coding agents can be added. Codex is planned next.
Handova itself makes no network requests. It copies files on your computer and starts your coding agent, which talks to its provider as it always does, under the account you picked.
The latest one from the folder you run it in. Add --pick to choose from a list, or --session <id> for a specific one. Run it from the same folder you started the tool in.
As many as you like. Personal and work are only examples: name them after clients, teams or anything else, and hand a conversation between any two.
Yes: handova handoff work personal. If you only carried on with one copy, handing it over again brings the older copy up to date.
Claude Code (the tool Handova supports today), and Node.js 20.16 or newer (or Bun). It's built and tested on macOS. Linux should work; Windows isn't supported yet.
No. It's an independent open-source project. It uses each tool's own commands for logging in and checking accounts, and changes nothing in the tools themselves.
Something else? Open an issue on GitHub.
Install it, then add your other accounts:
npm install -g handovahandova add work