Control, Scope & Route OpenRouter Keys Anywhere
Mint scoped API keys for your applications with granular model permissions, token rate limits, and full AES-256 encryption. Drop directly into any OpenAI-compatible client.
Curated Open-Source Frontier Models
The premier open-weights and reasoning models powered via OpenRouter upstream
DeepSeek R1
State-of-the-art open reasoning model utilizing reinforcement learning. Matches and exceeds OpenAI o1 on math, code, and logic benchmarks.
Meta Llama 3.3 70B Instruct
Meta's flagship open-weights model matching 405B performance at 70B efficiency. Outstanding instruction-following, analysis, and writing.
Qwen 2.5 72B Instruct
Premier open-source model for software engineering, multi-language coding, mathematics, and structured tool calling.
Mistral Small 24B Instruct
Ultra-fast European open-weights model designed for low-latency agentic loops, fast summarization, and rapid querying.
Military-Grade Security
Passwords hashed with bcrypt (salt cost 12), sessions tokenized with tamper-proof JWTs, anti-replay timestamp verification, and upstream credentials sealed with AES-256-GCM.
Granular Model Scoping
Restrict individual keys to single models (e.g. only DeepSeek R1 for reasoning tasks) to prevent unauthorized cost leakage and ensure model containment across projects.
OpenAI Drop-In Gateway
Zero code changes required in your applications. Point your existing Python, Node.js, LangChain, or cURL calls to /api/v1/chat/completions and pass your key.