Guide
Managed AI Models — Run Your Agent With No API Key
Every OpenClaw Launch plan includes managed AI models, so your agent can send its first message without you signing up for a provider account or pasting in an API key. Usage draws down an included credit balance instead. This guide explains what "no key needed" actually means, how managed models compare to bringing your own key (BYOK), how credits work, and how to switch models from your dashboard.
What "No Key Needed" Means
Normally, connecting an AI agent to a model means creating an account with a provider — OpenAI, Anthropic, Google, or an aggregator like OpenRouter — generating an API key, and pasting it into your agent's config. OpenClaw Launch skips that step by default. Deploy an agent, and it is already wired up to a managed model — one we host and route on your behalf. There is no provider sign-up, no key to copy, and no billing relationship with a third-party API to set up before your first message goes out.
This is not a trial gimmick layered on top of a BYOK-only product. It is one of two supported ways to run a model, alongside bringing your own key, and it is included in every plan.
Managed Models vs BYOK
OpenClaw Launch supports both approaches side by side, and you can switch between them at any time. Here is how they differ:
| Managed Models | BYOK (Bring Your Own Key) | |
|---|---|---|
| Who supplies the key | Nobody — we host/route the model | You — OpenAI, Anthropic, Google, OpenRouter, or any provider |
| Billing | Draws from your included credit balance | Billed directly by your provider account; credits aren't consumed |
| Setup friction | None — works the moment you deploy | Create a provider account, generate a key, paste it into your dashboard |
| Control | We manage provider routing and rate limits | Full control over which provider and account handles your traffic |
Neither option is a downgrade of the other — they are two ways to fund the same agent. Managed models get you running instantly; BYOK hands the billing relationship to you directly.
How Credits Work
When your agent runs on a managed model, each message draws down a credit balance rather than hitting a provider bill. That balance is visible on the Credits page, which shows how much you've used, how much remains, and your plan's limit for the current cycle.
- Included with your plan — an active subscription comes with a credit allotment that refreshes each billing cycle.
- Consumed per usage — every message your agent sends on a managed model uses a small amount of the balance. Nothing is deducted up front.
- Top-ups and other ways to extend usage — if you want to keep going after the included balance runs low, the Credits page offers a few options: connect a ChatGPT subscription, bring your own OpenRouter or other provider key, or refer a friend to earn more credit. See /pricing for plan details.
- BYOK doesn't touch the balance — once you connect your own provider key, usage is billed to that account instead, and your included credits stop being consumed.
Picking or Switching a Model
You are not locked into one managed model. The model selector for your agent lives on your dashboard, next to your instance — pick a different managed model there, or switch it from chat with the /model command. If you'd rather bring your own provider key, that's configured from the same dashboard, or from Credits where you can connect OpenRouter, a ChatGPT subscription, or another provider key.
When to Use Managed vs Bring-Your-Own-Key
Use a managed model if you want your agent working immediately, don't want to manage a separate provider account, or are just trying things out. It's the default for a reason — zero setup between deploying and chatting.
Use BYOK if you already have a provider account (especially a ChatGPT subscription you're already paying for), want usage billed directly to your own account instead of drawing down credits, or want a specific model/provider that isn't offered as a managed option.
How to Start
- Deploy an agent — Go to openclawlaunch.com, configure your agent, and deploy. It starts on a managed model automatically — no key required.
- Chat right away — Connect a messaging channel or use the web chat and send a message. Usage draws from your included credit balance.
- Watch your balance — Check Credits any time to see used, remaining, and limit for the current cycle.
- Switch any time — Pick a different managed model from your dashboard, or connect your own provider key from Credits to move to BYOK.
Related Guides
- OpenClaw BYOK — bringing your own provider key in depth
- Best Models for OpenClaw — choosing a model for your agent
- What Is OpenRouter? — how the OpenRouter aggregator works
- Is OpenClaw Free? — free/open-source vs managed pricing
- Pricing — plan details and included credits
Frequently Asked Questions
Do I need an API key to use OpenClaw Launch?
No. Every plan includes managed AI models, so your agent can chat the moment you deploy it — no provider account or API key required. Usage draws down an included credit balance instead. You can add your own provider key later if you prefer.
What happens to my usage on a managed model?
Each message your agent sends on a managed model consumes a small amount of your credit balance. You can see exactly how much you've used, how much remains, and your plan's limit on the Credits page.
What's the difference between managed models and BYOK?
With managed models, we host/route the model and usage is billed against your included credit balance. With BYOK, you supply your own provider key (or a ChatGPT subscription) and usage is billed directly to that account instead — your included credits aren't touched.
Can I run out of credits?
Yes — the included balance refreshes each billing cycle, but if you use it all before the cycle renews, you can extend usage from the Credits page: connect a ChatGPT subscription, bring your own provider key, or refer a friend to earn more credit. See /pricing for plan details.
How do I switch which model my agent uses?
The model selector is on your dashboard, next to your instance — switch models there or with the /model command in chat. You can move between managed models and your own provider key at any time.