AI Gateway Directory

AI API Gateways

A curated directory of AI API gateways and relays for OpenAI, Claude, Gemini, DeepSeek, Grok, Qwen, Midjourney and more — filter and compare by model, price, payment method, stability, and API compatibility.

40
Curated gateways

Scope

What's listed here

This page only shows the curated prepaid AI API relays, multi-model AI gateways, OpenAI-compatible proxies, and self-hosted API relays we maintain — so you can shortlist first, then visit each site for models, multipliers, payment, and compliance terms.

Model coverage

Tracks common API capabilities like OpenAI, Claude, Gemini, DeepSeek, Grok, Qwen, Moonshot, Doubao, Midjourney, Flux, Embedding, TTS, and Whisper.

Engineering features

Prioritizes flags that matter in production — OpenAI-API compatibility, streaming, function calling, log query, balance query, failure fallback, and self-hosting.

Risk framing

Prices, available models, and top-up methods change fast; the list only organizes information and never presents third-party stability, price, or data security as a guarantee.

Guide

How to choose a gateway

Choose a gateway as infrastructure, not just by the headline price.

Personal trial

Prioritize free quota, WeChat/Alipay top-up, small balances, and whether common models are all present. Don't top up too much at once.

Development & testing

Prioritize OpenAI-API compatibility, stable model names, clear error codes, balance query, and call logs.

Production

Prioritize SLA, status page, failure fallback, rate limiting, data-retention policy, enterprise billing, and backup providers.

Self-hosted gateway

Best for teams with ops capacity; use LiteLLM or One API to manage multiple official keys, grouped quotas, and logs.

Tutorial

API usage tutorial

Most OpenAI API relays are integrated by replacing the baseURL and apiKey, then filling in model names per the platform's requirements.

  1. Register a platform account and create an API key in the console.
  2. Check the docs for the base_url, model names, multipliers, and balance-query entry point.
  3. Use low-cost requests to test text, streaming, function calling, and error codes.
  4. For production, configure timeouts, retries, rate limiting, log redaction, and backup providers.
import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.AI_GATEWAY_API_KEY,
  baseURL: "https://gateway.example.com/v1"
});

const result = await client.chat.completions.create({
  model: "gpt-4.1-mini",
  messages: [{ role: "user", content: "Explain RAG in three sentences." }],
  stream: false
});

console.log(result.choices[0].message.content);

Risk

Risk notice

This site only organizes information about AI API relay services; it makes no guarantee about third-party platforms' stability, price, data security, or availability. Judge the risk yourself before use, and avoid uploading sensitive data, personal privacy, trade secrets, or core API keys to unknown platforms.

Sources

Sources and maintenance

The curated list is compiled by us across model coverage, payment methods, interface compatibility, pricing, and risk notes; public directories serve only as candidate supplements, and the displayed fields follow our normalized information.

FAQ

FAQ

Covers common questions about AI API relays, AI Gateway, AI Proxy, AI Relay, and OpenAI API relay.

What is an AI API gateway (relay)?
An AI API gateway usually wraps models like OpenAI, Claude, Gemini, and DeepSeek behind a unified interface, so you call many models with one key or one base_url. It can be a third-party aggregation platform or a gateway a team self-hosts.
What's the difference between an AI Gateway and an AI Proxy?
An AI Gateway emphasizes routing, fallback, logging, caching, rate limiting, and cost governance; an AI Proxy is more of a request-forwarding layer. Real products often blend the two, so check whether it offers observability, permissions, billing, and failure fallback.
Are AI relays safe?
Safety depends on the platform's credentials, log policy, key management, and the data you upload. On unknown platforms, avoid uploading IDs, business contracts, source-code secrets, customer data, or official API keys.
Could an AI API relay shut down and disappear?
Third-party relays can stop operating, delist models, or make balances non-refundable. Top up small amounts first, keep official APIs or backup providers for production, and export call records regularly.
How do I use an OpenAI API relay?
Usually you change the SDK's base_url to the address the relay provides and swap api_key for the relay's key. Model names, multipliers, and auth headers should be confirmed against the platform docs.
Can Claude API be relayed?
Yes, but confirm whether the platform supports Anthropic's native interface or Claude model names under the OpenAI-compatible format. Message format, tool calling, and streaming details can differ between the two interfaces.
Can Gemini API be relayed?
Yes. Some platforms offer an OpenAI-compatible interface for Gemini; others require Google's native format. Verify multimodal input, file upload, and safety settings separately.
Can OpenAI API be used from mainland China?
The official API's accessibility and payment rules vary by region. Third-party relays can lower the barrier to integration, but that isn't official authorization and doesn't replace a compliance review.
How is an AI API relay billed?
Common models include per-token, per-request, multiplier-based, plan-based, and self-hosted cost. When comparing, look at input price, output price, cache price, billing of failed requests, and refund rules together.
What is a multiplier?
A multiplier is the relay's conversion factor over the underlying model price — e.g. 1× is close to official cost, while a low multiplier may come from subsidies, bulk purchasing, or unofficial channels. The lower the multiplier, the more you should watch stability and risk control.
How do I choose a stable AI API relay?
Prioritize documentation completeness, a status page, operating history, failure fallback, log query, balance rules, and support responsiveness. For production, prepare at least two providers and set request timeouts and fallback models.
How does an AI relay differ from the official API?
The official API has clearer responsibility, terms, and billing; relays win on model aggregation, payment convenience, and a unified interface. They aren't replacements — important workloads should first evaluate official channels or a self-hosted gateway.

Submit

Submit a gateway

We welcome new AI Gateway, API Proxy, or self-hosted Relay projects and outage reports. Please include the site, docs, pricing page, model list, and risk notes.

See submission guidelines