AI API Providers
Bring your own key from 23 providers — or run on the credits included with every plan. Keys are encrypted at rest, and you can change provider after deployment without rebuilding your bot.
Run on a subscription you already pay for
No per-token API bill at all — your bot draws on your existing consumer plan.
ChatGPT Subscription
Run GPT-5.6 Sol, Terra and Luna on the ChatGPT Plus ($20/mo) or Pro plan you already pay for — zero API cost per token. Connect once with a terminal OAuth login.
Gateways and routers
One key, many models. The simplest way to bring your own billing.
OpenRouter
One key for the whole OpenRouter catalogue. Bringing your own skips platform credits and processing fees entirely — you pay OpenRouter directly.
OpenCode Go
Route chat through any model in the opencode.ai/zen/go aggregated catalogue using your existing OpenCode Go (Zen) plan.
Agnes
Free OpenAI-compatible multimodal gateway — chat, image and video generation with a 256K context window. Works on both OpenClaw and Hermes.
Nous Portal
Reach Hermes 4, Step 3.7 Flash (free for 30 days) and frontier models like Claude, GPT and Gemini through Nous Research’s own gateway.
Custom API Provider
Connect any OpenAI-compatible endpoint — Together AI, Mistral, a regional reseller, or your own self-hosted gateway. Base URL plus key, nothing else.
Direct model providers
Skip the middleman and talk to the model vendor directly.
Anthropic
Use Claude models directly on your own Anthropic account, without routing through OpenRouter.
OpenAI
GPT Image generation on OpenClaw — plus OpenAI chat and GPT Image 2 (low / medium / high quality) on Hermes.
Google AI (Gemini)
Use Gemini models directly and unlock voice-message transcription. Google AI Studio has a free tier.
xAI (Grok)
Route every Grok feature — chat, video and web search — directly through xAI.
DeepSeek
Send DeepSeek chat straight to DeepSeek instead of through a router.
Moonshot AI (Kimi)
Route chat directly through Kimi models on your own Moonshot account.
Groq
Ultra-fast Llama / Kimi / Qwen / GPT-OSS inference on Groq hardware.
Qwen / Alibaba DashScope
Route chat directly through Qwen models on Alibaba DashScope.
Search-grounded chat
Swap the chat model itself for one that answers from the live web, with citations.
Perplexity
Adds the Sonar models to your model dropdown — search-grounded answers with live citations. Sonar becomes your chat model and has no tool calling, so pick it when research matters more than agentic skills.
Image, video and music
Generation providers your bot can call straight from a chat message.
App integrations
Connect Gmail, Notion, Slack and hundreds more on your own account.
Composio
Run app integrations on your own Composio account and skip the daily action limit. Switching to your own key means reconnecting your apps once on the Integrations page.
How bring-your-own-key works
Three steps in the dashboard. The one exception is the ChatGPT subscription, which is connected with a one-time OAuth login — its setup guide walks through it.
Deploy your bot
Pick a model and launch. The bot runs on included credits from the first message — no key required to get started.
Paste your key
Open API Keys in the dashboard, paste the key for any provider on this page, and save. It is validated on the spot and stored encrypted.
Use it
Most chat and gateway keys add their models to the model picker, switchable any time. Media and integration keys light up the matching feature instead — image, video, music, app actions. Each card above says what its key unlocks.
Questions
Do I need my own API key to use OpenClaw Launch?
No. Every plan ships with credits included, and the free trial needs no key at all. Bringing your own key is optional — you do it when you want to pay the model provider directly instead of spending platform credits.
What is the cheapest way to run an AI agent 24/7?
Connect a ChatGPT Plus or Pro subscription and your bot runs GPT-5.6 at zero per-token cost — you are already paying the $20/mo. The other free routes are Agnes (free multimodal gateway), the free tiers at Google AI Studio and Groq, and the free-model router on OpenRouter.
Are my API keys safe?
Keys are encrypted at rest and only decrypted to configure your own container. You can remove a key from the dashboard at any time, which stops it being used immediately.
Can I use a provider that is not on this list?
Yes. The Custom API Provider slot accepts any OpenAI-compatible endpoint — a self-hosted gateway, a regional reseller, Together AI, Mistral, or anything else that speaks the OpenAI chat-completions format. You supply the base URL and the key.
Does bringing my own key remove platform fees?
For chat, yes — with your own OpenRouter key your messages skip platform credits and processing fees entirely. Hosting is still billed by your plan, because we run and maintain the container for you.
Can I mix providers?
Yes. A common setup is one provider for chat, fal or Replicate for image and video generation, and Composio for app integrations — those slots are independent and can be changed after deployment. The chat slot itself holds one model at a time, so choosing Perplexity Sonar means Sonar is the model answering you.
Bring your key, or use ours
Deploy an always-on AI agent in under 30 seconds. Add a provider key whenever you want.
Deploy Now