Skip to main content
Open source

multi-clawd

One Claude is never enough

An OpenClaw plugin that registers your extra Claude Code logins as first-class CLI backends. When your main account hits a usage limit, failover rolls to your second account on the same model — instead of dropping to a smaller one — with every tool, skill, and MCP bridge intact.

View on GitHubInstall
multi-clawd — a cyberpunk lobster with two extra robotic claws

Features

What it does.

First-Class Backends

Every account registers as a real CLI backend (claw2/…, claw3/…) — resolvable in model refs, fallback chains, and per-agent overrides. No API keys, no baseUrl hacks.

Same-Model Failover

Exhaust the account, not the model. A usage limit rolls to your next login on the same model — tier drops become the last resort instead of the default.

Full Harness on Every Hop

Each backend is a genuine Claude Code subprocess: native tools, skills, the OpenClaw MCP bridge, and native compaction all survive the hop across accounts.

Token Hygiene

Setup-tokens are read at launch and passed only through the child process environment. Never committed, never logged, never in your config in plain sight.

Self-Healing Registration

If the plugin loader hands over an empty config block, registration re-reads the resolved runtime config directly — a flaky pass can't silently disable your failover.

Isolated Logins

Each account lives in its own CLAUDE_CONFIG_DIR with its own token file, keeping sessions, settings, and auth cleanly separated per login.

How it works

4 simple steps.

1

Install the CLI

npm i -g @drakon-systems/multi-clawd, then multi-clawd update. The CLI installs the plugin with the right flags, offers the gateway restart, and runs a health check. Also on npm and ClawHub; no clone, no build step.

2

Run the Wizard

multi-clawd setup walks through accounts, isolated login dirs, token storage and the pool. It backs up openclaw.json before touching it, merges rather than overwrites, and supports --dry-run.

3

Sign In the Second Account

multi-clawd login claw2 launches the right Claude sign-in with the correct config dir and confirms which email landed. Tokens can resolve from your secret manager instead of a file on disk.

4

Verify the Routing

multi-clawd chain shows what actually serves each turn and flags anything bypassing the pool; doctor checks the whole install. Limits now roll across your accounts before any model-tier drop.

One chain, all your capacity

Your fallback chain stops meaning “smaller model” and starts meaning “next account”.

openclaw.json
"agents": { "defaults": { "model": {
  "primary": "claude-cli/claude-fable-5",   // main login
  "fallbacks": [
    "claw2/claude-fable-5",                 // 2nd login — same model
    "claw3/claude-fable-5",                 // 3rd login? go on then
    "anthropic/claude-opus-4-8"             // only NOW drop a tier
  ]
} } }

4

Claws

0

API keys required

100%

Harness intact

MIT

Licence

Frequently Asked Questions

What is multi-clawd?+
multi-clawd is an open-source OpenClaw plugin that registers additional Claude Code logins as first-class CLI backends. It lets your failover chain pool multiple Claude accounts — same model, next account — before ever dropping down a model tier.
Why does this need a plugin at all?+
OpenClaw's bundled claude-cli backend drives one Claude Code subprocess on a single login, and it won't hot-swap accounts mid-run. Plain config can't fix that: user-defined providers are treated as API providers that demand a baseUrl and API key. Plugin-registered backends get the same first-class treatment as the bundled one — that's the supported route, and the repo's DESIGN.md documents the dead ends so you don't have to rediscover them.
Does failover keep my skills and MCP tools?+
Yes — that's the whole point. Each account's backend is a genuine Claude Code subprocess with the OpenClaw MCP bridge, native tools, and skills attached. Stepping across accounts doesn't cost you the harness.
Is this within my provider's terms?+
multi-clawd is built for accounts you own. It runs the official Claude Code CLI under each login — no API spoofing, no shared credentials. You're responsible for using it within your provider's terms of service.
How many accounts can I add?+
As many as you own. Each entry in the accounts array becomes its own backend and provider prefix — claw2, claw3, and onwards.
How do I keep it up to date?+
One package installs as two halves. The plugin serves your turns and updates with multi-clawd update; the CLI is the command itself — doctor, chain, setup — and updates only with npm i -g @drakon-systems/multi-clawd@latest. Because update upgrades the plugin rather than itself, run that global install once by hand: from v1.6.0 the command warns you when it falls behind, but the warning ships inside the command, so an older one stays quiet. It matters because doctor and chain live in the CLI, and a stale command reports stale diagnostics about a perfectly current plugin. multi-clawd version shows both halves.
Does it pick models for me?+
No, and that's deliberate. It tells you when your provider ships a Claude model your chain doesn't mention, then stops. Which model belongs at the top of a chain is a cost, quality and behaviour decision that is yours — a plugin holding your credentials should never rewire your routing on its own.
What does it cost?+
Nothing. multi-clawd is MIT licensed and open source. If it saves your fleet from a single mid-task tier drop, it's paid for itself.

Grow an extra claw

Stop letting your second Claude account sit idle. Install multi-clawd and make your failover chain pool the capacity you already pay for.

Get Started on GitHub