ai-gateway

Configure unified AI model routing, provider failover, and cost tracking.

2|Updated Mar 29, 2026

Search Tags:#routing #llm #observability #failover #cost tracking #ai gateway

Ready to run

npx skills add https://github.com/NagyVikt/codex-plugins --skill ai-gateway-nagyvikt

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: ai-gateway
Source: https://github.com/NagyVikt/codex-plugins/tree/main/plugins/vercel/skills/ai-gateway
Command: npx skills add https://github.com/NagyVikt/codex-plugins --skill ai-gateway-nagyvikt

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires ai, @ai-sdk/gateway.

What problem does it solve?

This Skill solves the complexity of managing multiple AI providers by providing a single, unified API endpoint that handles model routing, provider failover, and cost attribution.

Core Features & Use Cases

Intelligent Routing & Failover: Automatically switch between providers like OpenAI, Anthropic, and Google if one is unavailable or rate-limited.
Cost & Usage Tracking: Attribute token usage to specific users or features using tags for granular budget management.
Use Case: A production SaaS application can use this to ensure high availability by failing over from GPT-5.4 to Claude-Sonnet-4.6, while simultaneously tracking costs per customer to prevent budget overruns.

Quick Start

Use the ai-gateway skill to configure a model routing chain that prioritizes openai/gpt-5.4 and falls back to anthropic/claude-sonnet-4.6 for a specific user.

Frequently Asked Questions about ai-gateway

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I set up LLM provider failover for my AI application?▼

LLM provider failover automatically reroutes traffic from an unavailable model to a backup, ensuring high availability. You configure a routing chain that prioritizes a primary provider and falls back to another for specific users.

What is the best way to track AI token costs per user in a SaaS application?▼

Tracking AI token costs per user requires attributing token usage to specific features using tags. This enables granular budget management and prevents overruns by monitoring consumption per customer.

Can I use Vercel AI Gateway with OIDC-based authentication for production traffic?▼

Vercel AI Gateway supports OIDC-based authentication for production traffic. It provides a unified API endpoint handling model routing, response caching, and audit logging across multiple LLM providers.

How does unified AI model routing handle rate limits across different providers?▼

Unified AI model routing handles rate limits by automatically switching between configured providers like OpenAI, Anthropic, and Google. If one provider rate-limits requests, the gateway reroutes traffic to an available fallback model.

Do I need per-user rate limiting for high availability LLM routing?▼

Per-user rate limiting is required for high availability LLM routing in production-grade applications. It works alongside provider failover and cost tracking to ensure stable performance and prevent budget overruns across users.