CatRouter gateway details

A third-party paid API relay that offers only Claude models and accepts both OpenAI and Anthropic request formats: the OpenAI-format endpoint is https://catrouter-api.com/v1 and the Anthropic-format endpoint is https://catrouter-api.com. If the primary address is unreachable, the backup https://catrouter.io works with the same key (add /v1 for OpenAI format). The pricing page, updated 2026-11-04, lists 11 models, including claude-opus-5-5, claude-sonnet-5-5, claude-haiku-5-5, and claude-opus-4-8. The website says it works with Claude Code, Codex, Cursor, Cherry Studio, and the openai SDK, and provides one-click import through CC Switch and Magpie.

Best for: For developers who mainly use Claude models, switch between clients such as Claude Code, Codex, and Cursor, and want to check public status data before a small top-up.

Current status

Test yourself

Latency reference

Depends on your deployment

/models check

Public pricing API /api/pricing; /v1/models requires a key

Last verified

2026-11-05

Status notes

The website, pricing page, public pricing API, and service status page were accessible; a keyless request to /v1/models returned 401. Availability, time to first token, and cache hit rate on the status page are collected by the platform itself; AI Beyond has not signed up, made paid requests, or tested latency. The top-up page requires sign-in, and Stripe payments and the USD-to-balance rate come from the submitter and remain unverified.

Pricing mode

Charged per token from a CNY balance: the official USD price is converted at CNY 1 = USD 1 and multiplied by 1.2. For example, the pricing page shows claude-sonnet-5-5 at CNY 2.40 input, CNY 12.00 output, and CNY 0.12 cache read per 1M tokens, and claude-opus-5-5 at CNY 4.80 input and CNY 24.00 output; claude-haiku-5-5 uses tiered pricing when a single input exceeds 100K tokens. Cache writes follow the official rules: 2x the input price for 1-hour caching and 1.25x for 5-minute caching; cache reads use the separate price on the pricing page. The submitter says top-ups go through Stripe (debit card, credit card, Alipay) with a USD 10 minimum and each USD 1 credited as CNY 6.68 of balance; the top-up page requires sign-in, and AI Beyond has not verified this.

Price tags

Pay-as-you-go

Free quota

Signup on the website needs only a username and password; the submitter says binding an email adds about CNY 3.9 of balance, which AI Beyond has not verified by signing up

Usually you create an API key in the platform console, then replace the SDK's baseURL, model name, and auth details per the docs. Test streaming, function calling, error codes, and rate limits before launch.

import OpenAI from "openai";

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

const response = await client.chat.completions.create({
  model: "claude-sonnet-5-5",
  messages: [{ role: "user", content: "Test one call through an AI gateway." }]
});

• Supported models: Claude.
• Capabilities: OpenAI-compatible.
• Documentation completeness: Medium; stability not tested.

• Not an official Anthropic service; official authorization has not been verified. The website says its upstream is "Claude Team accounts opened by a US company", and its caching notes refer to an "account pool"; AI Beyond has not verified these claims. Reselling team-account capacity may conflict with upstream terms of service and could stop working at any time. Requests pass through the platform's content safety checks: violating requests return 400, and keys with repeated violations are suspended. The service status page updates per-model availability, time to first token, and cache hit rate every minute, but the data is collected by the platform itself. The submitter says refunds and invoices are not supported, so start with a small top-up. The homepage does not link to terms of service or a privacy policy; avoid sending sensitive data. Usage is shown in the authenticated dashboard's usage log; this does not establish a public balance-query or log-query API.
• Third-party relay pricing, models, and availability can change, so re-check them periodically.

Is CatRouter compatible with the OpenAI API?

This entry is marked as supporting the OpenAI-compatible format, but the exact base_url, model names, and parameters still follow the official docs.

Is CatRouter ready for production?

Run a small-traffic load test, rate-limit test, and error-code test first, and confirm the billing rules; keep a backup provider for important workloads.