AI API relay — practical access notes

Actionable notes around “AI API relay” for teams that smoke-test before production.

What is an AI API relay?

An API relay unifies Claude, GPT, Gemini and more behind one OpenAI-compatible endpoint. 59API lets Claude Code, Codex and OpenAI SDKs share a single base URL and key — pay per token, no monthly lock-in.

Lower cost

Same model families at competitive usage rates.

Fast setup

Point base_url / BASE_URL at the relay and keep your SDK.

Multi-model

Claude + GPT + Gemini style access from one account.

Rollback-friendly

Keep official endpoints as a backup path.

How to evaluate a relay

Prefer protocol compatibility and smoke tests over marketing slides.

Dimension	What to check
OpenAI-compatible	SDK / Codex only need base_url + key changes.
Streaming	First-token latency and true SSE, not buffered dumps.
Errors	401/429 bodies are readable for ops.
Multi-client	Same key works for CLI, SDK, HTTP.
Topic	AI API relay
Host	3g.read.liuzhiwenhua.com

Config example

# AI API relay
export OPENAI_BASE_URL="https://59api.com/v1"
export OPENAI_API_KEY="sk-your-key"
# Map ANTHROPIC_* for Claude Code per relay docs if needed

FAQ

How do I connect?: Create a key in the console, set OPENAI_BASE_URL (or vendor-specific base) to the 59API endpoint, then run a minimal /models or chat test.
Is this the same as ChatGPT website login?: No. Web ChatGPT sessions are not API keys. Relays use sk- style keys and compatible HTTP APIs.
Which tools work?: Claude Code, Codex CLI, OpenAI SDK and most OpenAI-compatible clients after base URL mapping.

Start on 59API.com →