Pricing
Simple, transparent per-second billing. Pay only for what you use.
Agent VM
$0.033 / agent / hr
1 agent x 24hr = $0.79
Stopped Agent
$0.003 / agent / hr
1 stopped agent x 24hr = $0.07
Storage
$0.00018 / GB / hr
1 GB x 24hr ≈ $0.004
Agent Tokens
$3.60 / 1M input · $18.00 / 1M output
Willy AI assistant
| Resource | Rate | Example (24hr) |
|---|---|---|
| Agent VM | $0.033 / agent / hr | 1 agent x 24hr = $0.79 |
| Stopped Agent | $0.003 / agent / hr | 1 stopped agent x 24hr = $0.07 |
| Storage | $0.00018 / GB / hr | 1 GB x 24hr ≈ $0.004 |
| Agent Tokens | $3.60 / 1M input · $18.00 / 1M output | Willy AI assistant |
Free credits: New users receive $10.00 in free credits — enough to run 1 agent VM for ~300 hours (~12 days).
Agent Limits
- Each agent runs on a dedicated VM with allocated compute resources.
- A namespace can run multiple agent VMs simultaneously.
- Need more agents? Create additional agents within your namespace.
Billing Details
- Billing is per-second — you only pay for actual running time.
- All services are automatically suspended when your balance runs out.
- Add credits via credit card or on-chain stablecoin deposit (USDC / USDT) on Ethereum, Base, Arbitrum, Optimism, Polygon, and BSC.
- No minimum commitment — stop anytime, pay only for what you used.
Willy Tokens
- Tokens are consumed when you interact with your Willy agent via agent_chat or agent_invoke (MCP / A2A).
- Input tokens = your messages and context sent to Willy. Output tokens = Willy's responses. Output tokens are 5x more expensive than input tokens.
- Tokens are deducted immediately after each Willy response — billing is real-time.
- When your balance reaches zero, Willy requests are rejected (402) — no negative balance is allowed.
- Platform support queries (via ask_willy) do not consume tokens.
No refunds: All credit purchases and deposits are final and non-refundable. Since billing is per-second with no minimum commitment, you can stop services at any time. See our Terms of Service for details.
Data retention: When your balance runs out, all services are suspended. Agent VMs are stopped (not terminated) — add credits to resume. After a grace period, stopped VMs may be terminated and snapshots deleted.