Pay Per Successful Coding Task
No unmetered token waste. The inference compiler is free; monthly subscriptions deliver hosted serverless GPU inference with zero cold starts and 100% credit rollover.
Pro
For individual developers and engineers shipping daily production code.
- 4 concurrent coding agents
- ~12,500 MinTok 1 Pro tasks / mo
- MinTok 1 Max, MinTok 1 Pro, & MinTok 1 Flash access
- Background tasks & Git worktrees
- Monthly usage rollover (1 mo)
- No daily artificial token caps
Power
For developers running continuous parallel tasks and large codebase refactors.
- 8 concurrent coding agents
- ~25,000 MinTok 1 Pro tasks / mo
- Parallel agent task execution
- Priority GPU inference scheduling
- Larger task token budgets
- Monthly usage rollover
Max
Maximum throughput, dedicated warm pools, and priority inference queue.
- 16 concurrent coding agents
- ~50,000 MinTok 1 Pro tasks / mo
- Dedicated high-throughput GPU pools
- Highest task limits & long jobs
- Durable Objects shared team ledger
- GitHub Org sync & priority support
Ultra
For high-velocity teams running continuous multi-repo agent fleets.
- 40 concurrent coding agents
- ~131,000 MinTok 1 Pro tasks / mo
- Reserved high-priority GPU capacity
- Multi-repo distributed AST caching
- Webhook event streaming & routing SLA
- Priority 24/7 autonomous fleet support
Plan Feature Comparison
| CAPABILITY | PRO ($9.99) | POWER ($19.99) | MAX ($39.99) | ULTRA ($99.99) |
|---|---|---|---|---|
| Monthly Cloud Credits | $10.49 | $20.99 | $41.99 | $109.99 |
| Concurrent Agent Threads | 4 agents | 8 agents | 16 agents | 40 agents |
| MinTok 1 Max (1658 Elo) Access | Included | Included | Included | Included |
| MinTok 1 Pro (1638 Elo) Access | Included | Included | Included | Included |
| MinTok 1 Flash (1615 Elo) Access | Included | Included | Included | Included |
| AST Context Slicing Engine | Included | Included | Included | Included |
| Serverless GPU Scheduling | Standard Pool | Priority Queue | Dedicated Warm Pool | Dedicated Reserved GPUs |
| Shared Team Ledger | Individual | Individual | Up to 5 Seats | Unlimited Seats |
| Support SLA | Community | Email (24h) | Priority (4h) | 24/7 Dedicated Channel |
Looking for pure pass-through API token billing?
Access MinTok models directly via OpenRouter at exact pass-through prices: mintok/mintok-1-max ($0.42 / $1.32), mintok/mintok-1-pro ($0.085 / $0.24), and mintok/mintok-1-flash ($0.05 / $0.15).
Download MinTok Desktop Studio or Install CLI via cURL, npm, or Homebrew
The client-side inference compiler is completely free. Connect to your own serverless GPU fleet or active subscription.
Common Questions & Economics
How do monthly compute credits work?
Your subscription includes more credit than you pay (e.g., $10.49 included for $9.99/mo). Every API request or CLI task deducts the exact inference compute used. Unused credits automatically roll over to the following month.
Can I use MinTok with Cursor, Cline, or Aider?
Yes. MinTok provides an OpenAI-compatible endpoint at https://mintok.adstim.net/v1. Simply set your API key and baseURL in any existing tool.
Is my source code or prompt data stored or trained on?
Never for model training. MinTok operates as an in-flight optimization proxy to original upstream base models (OpenAI, Anthropic, Google Gemini, Mistral, DeepSeek). Neither MinTok nor the upstream API providers train on customer API inputs, prompts, or code. MinTok reduces tokens strictly in-memory during transit with zero disk storage of your repository files. Upstream model providers retain requests temporarily (0 to 30 days) solely for abuse and safety monitoring per their standard API terms. When using MinTok Engine with BYOK, your direct enterprise zero-retention agreements apply directly.
How does scale-to-zero work?
MinTok uses Cloudflare edge serverless infrastructure. When your agent fleet is idle, workers scale completely to zero so you never pay standby costs. Requests wake edge isolates with sub-second time-to-first-token.