AI coding CLI · cost optimizer
Mint routes every coding prompt to the cheapest capable model — automatically. No API keys to manage. No accounts to juggle. Just install and go.
cost reality
how it works
Mint reads your prompt and classifies it: write, debug, explain, review, or refactor. Each maps to an optimal price tier.
Simple coding tasks go to DeepSeek V3 at $0.27/1M. Complex reasoning goes to Sonnet. Only explicitly flagged tasks touch Opus.
Run mint usage to see exactly what you spent vs what you'd have paid on Opus-only. Share the report with your manager.
model routing
| Model | Provider | Input / 1M tokens | Output / 1M tokens | Axon routes when |
|---|---|---|---|---|
| DeepSeek V3 | deepseek | $0.27 | $1.10 | default write · fix · refactor |
| DeepSeek Coder | deepseek | $0.14 | $0.28 | auto targeted code generation |
| Qwen 2.5 Coder | openrouter | $0.40 | $1.20 | auto DeepSeek fallback |
| Claude Sonnet 4 | anthropic | $3.00 | $15.00 | auto explain · review · complex |
| Gemini 2.0 Pro | $1.25 | $5.00 | auto large context tasks | |
| Claude Opus 4 | anthropic | $15.00 | $75.00 | explicit only -m opus |
features
Classifies every prompt by task type and complexity. Maps to the cheapest model that can handle it. Fully configurable routing rules.
Scans your codebase, respects .gitignore, estimates tokens per file. Includes only what's relevant — fewer tokens = lower cost.
Real-time spend tracking per model. Shows predicted savings vs Opus baseline. Export reports for finance and team leads.
No API accounts, no key management, no rate limit headaches. Mint provides access to all 18 models under one subscription. Just install and start building.
pricing
quickstart
npm install -g mint-cli
click to copy
mint login
click to copy
mint
click to copy
mint usage
click to copy
Mint is in private beta. Join the waitlist and get early access — free while we're building. No credit card. No spam.