Choose your
AI model.
Access leading AI models through a single endpoint. Switch models anytime. $5 free credit included.
Four models. One proxy.
GPT-4o
Best for complex reasoning, multi-step tasks, and code generation. The default model for most users.
GPT-4o Mini
Fast and affordable for everyday tasks. Same quality for simple queries at a fraction of the cost.
Claude Sonnet 4.6
Excellent for writing, analysis, and nuanced conversation. Strong at following complex instructions.
Gemini Pro
Strong multilingual capabilities and broad knowledge base. Great for translation and international use.
How the LLM proxy works.
Single endpoint
Your agent calls one API endpoint. The proxy routes to the right model based on your configuration. No API key management needed.
Spend tracking
Every request is metered and tracked to your account. See exactly how much each model costs in your dashboard. No surprise bills.
Hot-swap models
Switch models at any time via your agent settings. No code changes. Try Claude for writing, GPT-4o for coding, Gemini for translation.
Use any model.
Already have API keys from OpenAI, Anthropic, or Google? Store them in the OmniClaw vault and use your own accounts directly.
BYOK lets you use any model your provider supports — including fine-tuned models, preview releases, and custom deployments. The proxy routes to your keys when configured.
Keys are stored encrypted (AES-256-GCM) in a per-account vault and written into your sandbox as an environment file only when a sandbox requests them. The API only ever returns key names.
# Store your OpenAI key in the vault
$ openclaw vault set OPENAI_API_KEY sk-...
# Configure agent to use your key
$ openclaw config set llm.provider openai
$ openclaw config set llm.model gpt-4o
$ openclaw config set llm.apiKeyVault OPENAI_API_KEY
# Your agent now uses your OpenAI account
✓ LLM configuration updatedStart with $5 free.
Try every model.
$5 on us.
Connect WhatsApp, pick a model, and start chatting. Switch anytime. No credit card required.