Top up.
Start working.
No subscription. One wallet — models, tools, and compute all deduct from it at the same flat rates, whichever package you start with.
Starter
Amy on your desktop — your machine is the compute.
one-time top-up · no subscription
Enough for 1 agent on your desktop, working all month
Top up & start- Desktop app (macOS & Windows) — compute free on your machine
- ~2M model tokens
- ~1,000 tool calls
- ✓Add cloud agents anytime — just top up
Cloud Agent
Everything you need for a working AI team.
one-time top-up · no subscription
Enough to keep 5 agents working 24/7 for a month
Top up & start- ~4,500 compute units
- ~5M model tokens
- ~2,500 tool calls
- ✓Desktop app included
Business
Scale up, with enterprise controls.
one-time top-up · no subscription
Enough to keep 10 agents working 24/7 for a month
Top up & start- ~9,000 compute units
- ~10M model tokens
- ~5,000 tool calls
- ✓Desktop app included
- ✓Single sign-on (SSO / SAML)
- ✓SLA + priority support
Package numbers are an example split — it's one wallet, spend it on any mix of models, tools, and compute. Top up again anytime.
Prices exclude VAT — applicable taxes are calculated at checkout.
One balance. Three things draw from it.
Top up a dollar wallet. As you use Amy, the cost deducts from it — same flat rates for everyone, nothing hidden.
Deducted per second while an agent's machine runs in the cloud. On your own machine, nothing is deducted.
Every model call (Claude, GPT, Gemini…) deducts from your wallet, priced per token.
Scraping, browser automation, and image or video generation deduct at the rates below.
Compute rate
| Size | Spec | Per day | ~Per month (24/7) |
|---|---|---|---|
| Small | 2 vCPU · 4 GB RAM · 20 GB SSD | $0.59 | $17.70 |
| Medium | 4 vCPU · 8 GB RAM · 40 GB SSD | $1.18 | $35.40 |
| Large | 8 vCPU · 12 GB RAM · 80 GB SSD | Coming soon | |
Disk — upgrade anytime in 10 GB steps, deducted from the same wallet
Charged while the disk is held, running or stopped. At 95% usage your machine pauses safely until you add more — the disk never actually fills.
Metered per second, only while running — the same flat rate for everyone, whichever package you top up with. A stopped machine pays only its disk rate.
Model rates
| Model | Input / 1M tokens | Output / 1M tokens |
|---|---|---|
| Claude Sonnet (closed, via OpenRouter) | $3.15 | $15.75 |
| GPT-4o (closed, via OpenRouter) | $2.63 | $10.50 |
| Kimi K2 | $0.58 | $2.31 |
| Qwen 2.5 32B | $0.71 | $1.26 |
| Qwen 2.5 Coder 32B | $0.71 | $1.26 |
| DeepSeek V3 | $0.28 | $1.18 |
| GPT-OSS 120B | $0.11 | $0.58 |
| GLM 4.7 Flash | $0.07 | $0.42 |
| Qwen 2.5 72B | $0.39 | $0.42 |
| Qwen 2.5 7B | $0.18 | $0.21 |
| GPT-OSS 20B | $0.04 | $0.16 |
Deducted per token at these rates. More models are available — the full catalog is in your dashboard.
Tool rates
| Tool | Unit | Rate |
|---|---|---|
| Web scrape / map | per call | $0.005 |
| Web search | per call | $0.01 |
| Web extract | per call | $0.03 |
| Web crawl | per page | $0.0005 |
| Browser session | per minute | $0.0011 |
| Browser bandwidth | per GB | $4.20 |
| Voice / speech (TTS) | per 1K characters | $0.0525 |
| Image generation | per image | from $0.0053 |
| Video generation | per video | from $0.0017 |
Tool calls use reserve-then-settle billing — we hold an estimated max, then charge actual usage at completion. Failed calls are never charged.
Want us to set it all up?
Our team designs, builds, and deploys assistants tailored to how your business actually works. Custom pricing, white-glove onboarding.
Book a consultationPricing questions
Is this a subscription?+
No. You top up once and use it until it runs out — then top up again whenever you like. Nothing renews, nothing is charged automatically.
Are the package numbers quotas?+
No — they're an example of what that amount typically buys. It's one wallet: spend it on any mix of model calls, tool calls, and compute. Run one heavy agent or ten light ones; the wallet doesn't care.
Can I run a big machine on a small package?+
Yes — no package limits your specs. Pick Small (2 vCPU · 4 GB · 20 GB disk), Medium (4 vCPU · 8 GB · 40 GB disk), or Large (8 vCPU · 12 GB · 80 GB disk, coming soon) on any package; each size has its own flat day-rate, the same for everyone. A bigger machine simply deducts faster, so you'll top up sooner. That's the whole rule.
Do bigger packages get better rates?+
Rates are flat and identical on every package — you're simply starting with more runway. The $200 Business package additionally unlocks SSO/SAML, an SLA, and priority support.
Does a cloud agent cost money even when I'm not using it?+
Its machine deducts compute per second while running — from $0.59/day for Small up to $1.18/day for Medium. Stop the machine and you pay only the disk rate (about $2 per 10 GB per month) to keep your data. Agents on your own machine never deduct compute.
Can I upgrade my disk?+
Yes — add storage in 10 GB steps, about $2 per 10 GB per month, deducted from the same wallet. If your disk reaches 95% full, the machine pauses safely — nothing is lost, the disk never actually fills — and adding storage resumes it right away.
What happens when my balance runs out?+
Cloud machines pause automatically so you're never surprised — nothing is deleted, and topping up resumes them immediately. Agents on your own machine simply stop making model and tool calls until you top up.