Introduction

What is XAI Router?

XAI Router is a high-performance, high-availability AI API Router and management platform (including a full suite of multi-tenant management APIs) that integrates transparently at the application layer, also known as the Yugong Router. It is designed to solve the core pain points encountered when using native AI services (like Deepseek, OpenAI, Anthropic, etc.) directly, such as stability, concurrency limits, key management, cost control, and multi-tenancy management.

If you think of model providers like DeepSeek, OpenAI, and Anthropic as "power plants," then XAI XAPI is a business engine that combines the "national grid (for highly reliable transmission) + regional power distribution companies (for the multi-tenant system) + a smart meter billing system (for granular management)" all into one.

Service Models

• Private Deployment: XAI Router [1]: A single-tenant architecture. System administrators add their AI APIs, and can then create sub-accounts to share and manage those AI resources. This model is for large and medium-sized enterprises, financial institutions, healthcare organizations, and government agencies with stringent requirements for data sovereignty, security, and compliance. We provide a complete on-premises deployment solution, ensuring your data never leaves your domain and all compliance boundaries are respected.
• Cloud Service: XAI Control [2]: A multi-tenant architecture. Upon registration, you gain ownership of an independent AI router, allowing you to create sub-accounts and manage shared AI resources. This service is designed for global developers, offering convenient, high-availability AI API services that enable seamless migration and rapid deployment.

Whether you need the convenience of the public cloud or the security of a private deployment, we can provide the professional solution that best suits your needs.

Core Features

• Unified API Endpoint and Routing: All model calls go through a single domain, with the router intelligently routing requests to the optimal upstream provider.
• Intelligent Key Rotation and Distribution: Real-time monitoring of key balances and health status, with automatic rotation, circuit breaking, and recovery within the key pool to avoid single-point rate limiting.
• Rate Limiting, Authentication, and Usage Statistics: Supports TPM/TPH/TPD/RPM/RPH/RPD rate limiting; control policies like IP whitelisting and model whitelisting; real-time usage reporting and a visual dashboard.
• Automatic Failover: Upstream errors (timeouts, 429s, 5xxs) trigger millisecond-level circuit breaking, with automatic retries and switching to healthy nodes, completely transparent to the client application.
• Model Mapping: Supports model alias mapping at both the system and account levels. For example, map gpt-3.5-* to gpt-4o-mini or deepseek-chat to deepseek-coder, allowing you to smoothly upgrade or downgrade models without changing existing code.
• Ultimate Stability: Through our built-in key pool and intelligent routing algorithms, we effectively shield you from common upstream errors like 429 and 5xx from services like OpenAI, maximizing the success rate of your business requests.
• Seamless Compatibility: While supporting native APIs, we are also compatible with the OpenAI and Anthropic API specifications. You only need to change the base_url and api_key in your existing application to switch seamlessly, with no code logic modifications required.
• Powerful Account System: Provides flexible, multi-level sub-account management. You can create independent sub-accounts for team members or customers and finely control their rates, quotas, rate limits (RPM/TPM), and available models.
• Granular Cost Control: All consumption details are trackable, from a global overview to itemized records, making your costs clear and controllable. Supports pay-as-you-go billing with prices consistent with official model rate cards.
• High Performance: The core service logic runs entirely in memory, ensuring ultra-low latency and excellent concurrent processing capabilities.
• Zero-Trust Security Design: In the multi-tenant XAI Control system, core data (XAI API Keys and upstream AI API Keys) is stored with cryptocurrency-grade end-to-end encryption. We implement a "user-as-key" decentralized security policy—your data can only be unlocked by your exclusive key, fundamentally eliminating the risk of data leakage.

How It Works

Whether you are an independent developer, a startup team, or an enterprise user, XAI XAPI can provide you with a stable, efficient, and controllable AI API service, allowing you to focus more on business innovation.