Explore
$0
Create and test on your device.
- Local workspace and source editing
- Browse working examples
- Connect your own services
Free public demo · Future paid plans
Build freely.
Pay for what runs.
The device demo is free and needs no account. Cloud accounts require verified email and administrator approval. The paid plans below are proposals, not available for purchase.
Studio
Draft$25 / month
One subscription for a shared team wallet.
- 20 service credits per paid month
- Credits shared across your team
- Model and spending limit before every run
- Manage renewal through Stripe
Pay as you go
$10 to start
Prepaid usage when you need more room.
- $10, $25, and $50 credit packs
- No recurring commitment
- Purchased credits have no expiry
- Available when merchant billing is configured
One service credit = $1 of quoted usage. Unused Studio credits carry forward; the current ledger does not reset or expire them. Credits are internal service balances, non-transferable and not crypto tokens. Provider charges for your own keys and app hosting are separate. No unlimited AI, unmetered agents, or paid feature promises are included in this proposal.
Transparent usage
Know what every model costs.
Proposed retail rates are the provider’s standard global text API price plus 25%. This is our draft margin, not an industry standard. Prices below are USD per one million tokens, checked 2026-10-14.
| Model | Input | Cached input | Output |
|---|---|---|---|
| GPT-6.1 Sol ↗OpenAI · standard context | $2.50API $2.00 | $0.125API $0.10 | $12.50API $10.00 |
| Claude Sonnet 5.5 ↗Anthropic · standard context | $2.50API $2.00 | $0.125API $0.10 | $12.50API $10.00 |
| Claude Opus 5.5 ↗Anthropic · standard context | $5.00API $4.00 | $0.25API $0.20 | $25.00API $20.00 |
| Grok 4.7 ↗xAI · standard context | $2.50API $2.00 | $0.625API $0.50 | $7.50API $6.00 |
Cache writes, long context, and other charges
- GPT-6.1 Sol: Proposed cache writes $3.125/MTok. Over 272,000 input tokens: $5.00 input / $0.25 cached / $18.75 output per MTok for the whole request; cache writes $6.25/MTok.
- Claude Sonnet 5.5: Proposed cache writes $3.125/MTok (5-minute). One-hour writes $5.00/MTok.
- Claude Opus 5.5: Proposed cache writes $6.25/MTok (5-minute). One-hour writes $10.00/MTok.
- Grok 4.7: No separate cache-write rate listed. At least 200,000 input tokens: $5.00 input / $1.25 cached / $15.00 output per MTok for the whole request.
Reasoning tokens are billable output. Tools, sandbox execution, image/audio/video, regional processing, and faster service tiers have separate costs. This calculator excludes them. The draft does not change deployed billing rates: every hosted run uses its server-issued quote and spending ceiling. Cache discounts are only available where the runtime meters cache usage separately.
Estimate a text request
Draft retail estimate, uncached input. Multi-agent work adds usage across every agent and review round.
Estimated service credits
Standard-context rates.
Bring your own intelligence
Connect provider keys, a Hugging Face deployment, or an approved local model endpoint. Those provider bills stay with you. Hive runs use an explicit budget and collect usage from each agent; a local model does not imply free hosting or unlimited compute.
How we arrived at the draft
Current products combine subscriptions and usage in different ways. Cursor lists $30/month Premium with third-party API usage billed separately, while Lovable pools credits across workspace members. Our proposed $25 Studio plan and 25% API markup are an independent commercial choice. They still need merchant approval before sale.