← Back to all guides

Gemini CLI API relay: is 59API right for solo developers

The problem: too many accounts, too many bills

If you are a solo developer, you have felt this. You want Claude for one task, GPT for another, DeepSeek for a cheap batch job, and Gemini CLI for something else entirely. Before you write a single line of code you are juggling separate sign-ups, separate keys, separate invoices and separate prepaid minimums. Some months you burn through a subscription; other months you pay for nothing at all. And every time you switch models, you rewrite your client code again.

That is the situation 59API was built for. It is an independent third-party AI API relay: one account, one API key and one API host, https://api.59api.com, for Claude, GPT, Gemini, DeepSeek and other models. You top up once and spend across families.

Billing is prepaid balance charged per request by real token usage. There is no monthly fee, and balance never expires. A quiet month costs you nothing; a busy month costs exactly what you used. Top-ups start at ¥10 and go up to ¥1000, paid by WeChat Pay or bank card, and invoices are available when you need to file one.

For a solo developer, that is the whole appeal in one sentence: one key, no subscription, pay only for what you actually run.

What a solo workflow looks like on 59API

Picture a hypothetical solo developer. In the morning they are pair-programming in Claude Code on a hard refactor. After lunch they use Gemini CLI for a research-heavy task. In the evening they queue a long, cheap batch through an OpenAI-compatible client. On 59API, all three run off the same balance and the same key.

Connecting is a matter of environment variables. For Claude Code or Claude Desktop, set ANTHROPIC_BASE_URL=https://api.59api.com and ANTHROPIC_AUTH_TOKEN=YOUR_59API_KEY. Claude models work only through the genuine Claude Code or Claude Desktop client, Claude Code 2.1.285 or newer, never through the OpenAI SDK, cURL or other clients.

For Gemini CLI, set GOOGLE_GEMINI_BASE_URL=https://api.59api.com and GEMINI_API_KEY=YOUR_59API_KEY. Gemini models use Google's native protocol only; they are not OpenAI-compatible and do not work on /v1/chat/completions.

For Codex, OpenAI SDKs, cURL and OpenAI-compatible clients such as Cursor, Cherry Studio, Cline and OpenCode, the base URL is https://api.59api.com/v1 with the chat endpoint https://api.59api.com/v1/chat/completions. Those clients carry GPT, DeepSeek and other non-Claude, non-Gemini models only. One key, three protocols, no rewrites.

How the pricing behaves when you are one person

Prices are per 1 million tokens, billed in CNY, with USD figures shown at ¥7.2 = $1. What matters for a solo budget is the shape of the cost: input is cheap, output is the real spend, and cache read is close to free.

Take Claude Sonnet 5.5 (claude-sonnet-5-5): input $1.11 (¥8), output $5.56 (¥40), cache read $0.0972 (¥0.7) per 1 million tokens, against an official list price of $3 / $15. Or Claude Haiku 4.5 (claude-haiku-4-5-20251001): input $0.3472 (¥2.5), output $1.67 (¥12), cache read $0.0278 (¥0.2).

For GPT work, GPT-6.1 Sol (gpt-6.1-sol) is input $0.1847 (¥1.33), output $0.9264 (¥6.67), cache read $0.0375 (¥0.27), with an official list price of $2 / $10. DeepSeek V4 Flash (deepseek-v4-flash) runs input $0.2778 (¥2), output $1.11 (¥8), cache read $0.006944 (¥0.05).

Two worked examples show the effect. GPT-6.1 Sol at 5M input plus 1M output tokens without cache is $1.85 (¥13.32) on 59API versus $20 (about ¥144) at official list price, about 9% of list price. Claude Sonnet 5.5 at 3M input plus 0.5M output tokens without cache is $6.11 (¥44) versus $16.5 (about ¥118.8), about 37% of list price.

When prepaid beats a card for one person

59API also offers optional cards, but a card is a different tool. A card has its own key, can only call that family's models and is paid only from the card: it never falls back to wallet balance, and a balance key never draws on a card. Unused quota does not roll over, and quota and caps are in official-price USD.

For a solo developer with uneven output, prepaid balance is usually the honest fit. You are not pre-committing to a 30-day window you might not fill. If you do have a known sprint, a card can make sense. The weekly GPT card, 【周卡】G-PRO $1500, is $40.56 (¥292) for 7 days with quota $1500 and weekly cap $1500, draining at official price × 2.4, so about $625 of usage at official list prices.

For a longer push, the Claude card C-PRO 入门版 is $83.19 (¥599) for 30 days, quota $1300, weekly cap $500, draining at official price × 2.4, so about $541.67 of usage at official list prices. Both are optional. Nothing forces you into one.

The rule of thumb: predictable heavy use for a fixed stretch points at a card; everything else points at prepaid balance.

Who 59API suits, and who it does not

59API fits a solo developer who wants to test or ship across model families without maintaining several accounts, who prefers paying per request over a monthly fee, and who values a balance that never expires. It fits the developer who lives in Claude Code and Gemini CLI and occasionally drops into an OpenAI-compatible client for GPT or DeepSeek. And it fits anyone who wants invoices and a low barrier to entry, with top-ups from ¥10.

It is less of a fit if you need Claude outside the genuine Claude Code or Claude Desktop client, or Gemini outside Gemini CLI. Those limits are real: Claude is not available through the OpenAI SDK, cURL or third-party clients, and Gemini is not OpenAI-compatible.

It is also less of a fit if you want a flat monthly subscription instead of usage-based billing, since 59API charges per request by real token usage. And if you only ever use one model family, weigh the optional cards against prepaid balance before deciding.

If the first three paragraphs sounded like your week, 59API is worth a look. Start with a small top-up, point your client at https://api.59api.com, and see how far your tokens go.

Docs