If you use ClawCloud's managed AI credits addon, your instance gets a monthly pool of AI credits. Your bot uses these credits whenever it responds to a message through OpenRouter. This guide covers how credits work, how to check your usage, and how to stretch your budget.
What are AI credits?
AI credits are a dollar-denominated allowance for AI API usage. When your OpenClaw bot processes a message, the API charges based on tokens (roughly 4 characters per token). That charge comes out of your credit pool.
Each addon tier gets a different monthly amount:
| Addon Tier | Monthly Price | AI Credits |
|---|---|---|
| Small | +$9/mo | $8/mo |
| Medium | +$28/mo | $25/mo |
| Large | +$66/mo | $60/mo |
| XLarge | +$110/mo | $100/mo |
Credits reset at the start of each billing cycle. Unused credits don't roll over.
Check your credit usage
Open your ClawCloud dashboard and look at the credit bar under your instance. It shows dollars used out of your total allowance, with color coding:
- Blue — 75% or below, plenty of credits remaining
- Orange — above 75%, approaching the limit
- Red — exhausted (paid-model responses stop until you switch manually)
What happens when credits run out
Your bot stops responding on paid models once credits are exhausted. The dashboard shows an exhaustion warning and links to a quick manual switch flow for free models.
Free models are available in the catalog, and you can switch with /model to restore service. Free models are useful for routine tasks but quality and speed are lower than paid models.
When credits reset on your next billing date, your paid model works again.
Which free models are available?
ClawCloud includes free models from multiple providers via OpenRouter:
| Model | Provider |
|---|---|
| Llama 3.3 70B Instruct | Meta |
| Qwen3 Coder | Qwen |
| Mistral Small 3.1 24B | Mistral |
| Gemma 3 27B | |
| DeepSeek R1 | DeepSeek |
These models cost nothing against your credits. They handle everyday conversations well, though they won't match flagship models like Claude Sonnet or GPT-4.1 on complex reasoning tasks.
How many messages do you actually get?
It depends on the model. Cheaper models stretch your credits much further:
| Model | Lite ($8) | Pro ($25) | Max ($60) |
|---|---|---|---|
| GPT-4.1 Mini | ~11,400 | ~35,700 | ~85,700 |
| Gemini 2.5 Flash | ~8,900 | ~27,800 | ~66,700 |
| GPT-4.1 | ~2,350 | ~7,350 | ~17,600 |
| Claude Sonnet 4 | ~1,300 | ~4,200 | ~10,000 |
A typical Discord or Telegram bot handles 20–100 messages per day. At 50 messages/day (1,500/month), even the Lite plan handles that comfortably on GPT-4.1 Mini or Gemini Flash.
Tips to get more from your credits
Pick the right model for the job. GPT-4.1 Mini handles most daily conversations just as well as the full GPT-4.1, at roughly 5x lower cost. Save the premium models for when you need longer reasoning or complex tasks.
Keep your system prompt short. The system prompt is sent with every message. A 2,000-token prompt costs you tokens on every single exchange. Keeping it under 500 tokens saves real money over a month.
Add "be concise" to the system prompt. AI models default to verbose responses. Shorter answers mean fewer output tokens, and output tokens are 3–5x more expensive than input tokens.
Switch models from chat. Use the /model command to swap between models on the fly. Use gpt-mini for routine questions and switch to sonnet or gpt when you need more firepower. See the model switching guide for details.
Model defaults
All plans default to cost-efficient models: GPT-4.1 Mini when you select GPT, Gemini 2.5 Flash when you select Gemini, and Claude Haiku 4.5 when you select Claude. These models give you more messages per dollar while still being very capable.
You can switch to any of the 94 curated models in the catalog using /model. The default just determines your starting point. See the model switching guide for details.
Bring your own key (BYOK)
If you'd rather skip credits entirely and pay your AI provider directly, switch to BYOK (bring your own key) mode during deployment. You'll provide your own API key from OpenAI, Anthropic, or Google, and ClawCloud won't track or limit your usage.
BYOK makes sense if you expect heavy usage or want to use a model not available through OpenRouter.
Troubleshooting
Credit bar isn't updating — Credit usage is polled every 5 minutes. Wait a few minutes and refresh the dashboard. If it stays stuck, contact help@clawcloud.sh.
Credits reset but my model didn't change back — When credits reset on your billing date, ClawCloud restores your original paid model automatically. If you manually switched to a free model via the dashboard, that persists until you change it back.
I want to check exact usage numbers — The dashboard credit bar shows usage against your plan limit. For more detail, you can check your OpenRouter usage via the OpenRouter dashboard if you have access to the sub-key (managed mode keys are provisioned automatically, so this typically isn't available).
Deploy with Managed AI Credits