👋 Welcome!
We provide access to Claude Code CLI based on a compatible API solution for use in popular development environments.

💻 The service is suitable for use in tools and IDEs where model connection via a compatible API interface is required.
Various working environments are supported, including VS Code, Cline, Cursor, OpenCode, Antigravity, OpenClaw, and other compatible solutions.

📦 After purchase, you will receive:
— an API key for connection
— detailed setup instructions
— examples of use in popular development environments
— access to a dashboard for tracking requests and usage

🧩 In some configurations, models may be displayed via Kiro provider ID. If necessary, you can independently review the official information about Kiro in open sources.

⚙️ The main use case is development- and coding-related tasks:
— work in CLI
— use in code editors
— connection via extensions
— use in dev environments and coding tools

🚫 For public chatbots, mass API integrations, high-load services, and constant parallel workloads, the access may be less suitable, since request limitations may occur in such scenarios.

📱 Terms of use:
— the key can be used on different devices
— simultaneous activity in multiple places may lead to limitations
— the usage format предполагает личную работу, а не коллективный доступ

🕒 Warranty periods:
— up to 20M tokens — 15 days from the date of purchase
— up to 50M tokens — 10 days from the date of purchase
— for a larger volume (100M+) — 3 days from the date of purchase

🔁 Keys may remain active for a longer period, however, the guaranteed period is limited to the timeframes specified above. Therefore, it is better to start using the key immediately after purchase.

💬 Refunds:
If the service did not suit you, write to us within 12 hours after purchase.
With minimal key usage, the refund request may be considered individually.

📊 How to verify the quality of access:
We recommend comparing the model’s performance on your real tasks, as well as on open benchmark sets from GitHub.
Small differences in responses are acceptable, since model generations may vary from one run to another.