OpenClaw × Byoky

Run Claude Code in OpenClaw — for free.

Claude Code
Codex
Gemini CLI

Grab a free token gift or use your Claude Pro/Max subscription.
Zero extra cost. Keys never exposed.

Start Setupor
Requires the Byoky extension v0.9.0+. Available now on Firefox and Android and iOS.
Two ways to run OpenClaw for free

Token gift from the token pool

Anyone can gift token access on the Byoky token pool. OpenClaw runs entirely on the gifter's budget — capped, revocable, and proxied through their wallet so their key never leaves their machine.

No card, no signup, no provider account.

Your Claude Pro/Max subscription

Already paying for Claude Pro or Max? Connect it to Byoky with a setup token and OpenClaw runs on the same subscription credits you already use in claude.ai. No API credits, no per-token billing.

Anthropic-only — uses your existing plan.

1

Install the Byoky wallet

The wallet holds API keys (yours or gifted) and proxies every request.

Pick the version for your machine:

Open the wallet, set a password, and you're ready. No accounts, no email.

2

Add a credential

Three options — pick whichever fits.

Free token gift — grab one from the Token Pool. No API key needed.
Claude Pro/Max setup token — run claude setup-token in your terminal and paste it into Byoky.
Your own API key — add it directly in the wallet.
3

Install Byoky for OpenClaw

One command installs the plugin and its bridge dependency.

openclaw plugins install @byoky/openclaw-plugin

The plugin registers all 13 Byoky providers with OpenClaw and pulls in @byoky/bridge, the tiny local HTTP proxy that lets OpenClaw talk to your wallet. The native messaging host is registered on the next step when you first connect — no extra command.

4

Connect OpenClaw to your wallet

One command connects every provider you have in the wallet.

openclaw models auth login --provider byoky

First run: OpenClaw asks to register the native messaging host (press Enter to accept), then opens your browser so the wallet can approve the connection. Every provider you have in the wallet — Anthropic, OpenAI, Gemini, whatever — gets configured in one shot.

Want just one provider? Use --provider byoky-anthropic (or byoky-openai, byoky-gemini, byoky-xai, etc.) instead.

Verify the bridge:

curl http://127.0.0.1:19280/health
# → {"status":"ok","providers":["anthropic","openai",...]}

Subsequent runs skip the browser tab if the bridge is already live — they just re-use the session.

5

Pick your default model

OpenClaw's onboarding picks a non-byoky default. Point it at a byoky model so chat works out of the box.

openclaw models set byoky-anthropic/claude-sonnet-4-6

Swap the model for whichever you actually have a key or gift for: byoky-openai/gpt-4.1, byoky-gemini/gemini-2.5-pro, byoky-xai/grok-3, etc. openclaw models list --all shows every option.

6

Use OpenClaw

That's the whole setup. Run OpenClaw as you normally would — every LLM call routes through your wallet.

Inside OpenClaw, the /byoky command shows bridge status and lists every connected provider:

/byoky

Token usage is tracked in your wallet's Sessions view. If you're using a gifted credential, the gifter's budget ticks down in real time and stops you cleanly when it hits zero.

If something doesn't work

Chat hangs on "..." forever

The bridge session got out of sync with the extension (happens after a browser restart or wallet re-lock). Re-run openclaw models auth login --provider byoky — plugin 0.7.7+ detects this and forces a fresh wallet re-pair.

"No API key found for provider \"openai\""

OpenClaw is still pointing at the direct openai plugin's default. Run openclaw models set byoky-anthropic/claude-sonnet-4-6 (or another byoky model) to fix. Optionally openclaw plugins disable openai if you never plan to use it directly.

"invalid x-api-key" from Anthropic (or OpenAI)

The key saved in your wallet is wrong or revoked. Grab a fresh one from the provider's console and paste it in, or use a free gift from the token pool instead.

I added a key to the wallet but the chat still says "provider not available"

The bridge snapshotted your providers before you added the key. Re-run openclaw models auth login --provider byoky to refresh the session with the updated provider list.

Fixed the bad credential, but OpenClaw still won’t use that provider

After a credential fails (bad key, expired gift, rate limit), OpenClaw puts the provider profile in an in-memory cooldown that persists until the gateway restarts. Restart it with openclaw gateway restart and try again.

Anthropic keeps returning 429 rate_limit_error on a gift

The gifter\u2019s upstream key is being throttled by Anthropic (the limit is on their account, not yours). Try a different Anthropic gift, switch the chat model to a non-Anthropic byoky provider, or wait for the window to clear (usually 1 hour).

Inside OpenClaw, /byoky prints bridge status and connected providers.openclaw logs --follow shows the raw errors if you need to dig deeper.

Available providers

The plugin registers all 13 Byoky providers with OpenClaw:

ProviderOpenClaw IDModels
Anthropicbyoky-anthropicClaude Opus 4, Sonnet 4, Haiku 4.5
OpenAIbyoky-openaiGPT-4.1, o3, o4-mini, GPT-4.1 Mini
Google Geminibyoky-geminiGemini 2.5 Pro, 2.5 Flash
xAIbyoky-xaiGrok 3, Grok 3 Mini
DeepSeekbyoky-deepseekDeepSeek V3, R1
Mistralbyoky-mistralMistral Large
Groqbyoky-groqLlama 3.3 70B
Coherebyoky-cohereSet model manually
Perplexitybyoky-perplexitySet model manually
Together AIbyoky-togetherSet model manually
Fireworks AIbyoky-fireworksSet model manually
OpenRouterbyoky-openrouterSet model manually
Azure OpenAIbyoky-azure_openaiSet model manually

How it works

OpenClaw → HTTP → Bridge (localhost:19280) → Native Messaging → Extension → LLM API
                                                                ↑
                                                       Keys live here. Always.
  1. The plugin registers each Byoky provider in OpenClaw, pointing at http://127.0.0.1:19280/<provider>.
  2. OpenClaw sends an LLM request to the local bridge.
  3. The bridge relays the request to your Byoky extension via Chrome native messaging.
  4. The extension injects the real API key (yours or a gifted one) and calls the provider.
  5. The response streams back through the same path.

If you're using a gifted credential, requests are relayed once more — through the gifter's extension, which holds the real key and enforces the token budget. You see streaming tokens; the gifter sees a usage counter ticking down.

Ready to try it?

Grab a free token gift from the token pool and you'll be running Claude or GPT inside OpenClaw in under five minutes.