Zerouter
AI model API gateway — one endpoint for many providers
In their words
AI model API gateway — one OpenAI-compatible endpoint for many providers.
Zerouter is an AI API gateway that lets a single API key call Claude, GPT, Gemini, Grok and Kimi models through OpenAI and Anthropic compatible endpoints. It operates on a prepaid credit system with token pricing up to 93 percent below official list prices and charges no monthly subscription fees.
Endpoints and Routing
The gateway provides compatible API channels on POST /v1/chat/completions, Anthropic on POST /v1/messages, OpenAI on POST /v1/responses, Google on POST /v1/chat/completions, xAI on POST /v1/responses, and Moonshot AI on POST /v1/messages. The model ID named in the request runs directly without prompt rewriting or model substitution.
Client Integrations
-
Claude Code: configure ANTHROPIC_BASE_URL and provide the key in ANTHROPIC_AUTH_TOKEN.
-
Codex CLI: add a model providers block to ~/.codex/config.toml with the base URL and environment key.
-
Cursor: enable base URL overrides and paste the key alongside the model identifier.
Fact file
Developer tool- API
- Claude Code
- Codex CLI
- Cursor
- Models
- 36 models across 8 families
- Protocols
- 3 protocols
What is Zerouter?
Endpoints and routing
Requests route through a single base URL located at api.zerouter.ai. The gateway accepts calls via multiple paths depending on the provider and model. The POST /v1/chat/completions route handles Google, Gemini, Qwen, DeepSeek, and GLM models. The POST /v1/messages path runs Anthropic and Moonshot AI models. The POST /v1/responses endpoint serves OpenAI and xAI models. Each request executes the exact model identifier supplied in the payload without prompt rewriting or automatic model substitution, ensuring responses match official upstream behavior.
Client configurations and integrations
Connecting existing development tools requires updating base URLs and authentication environment variables. Claude Code uses the ANTHROPIC_BASE_URL setting combined with the API key assigned to ANTHROPIC_AUTH_TOKEN. Codex CLI reads a model providers block placed inside the configuration file located at ~/.codex/config.toml, mapping the base URL and the environment key variable. Cursor users enable base URL overrides directly within the application settings before pasting the credential alongside the target model identifier. Any standard OpenAI Software Development Kit can point to the gateway by overriding the default endpoint URL.
Billing and token pricing
Operations run on a prepaid balance model where funds deduct per input, output, and cache token consumed during inference. Pricing varies by model family, offering rates positioned below standard public list prices. The system logs every call with its exact token count and cost, making usage tracking visible directly in the dashboard console. Crypto payment channels incur no processing fees when topping up credits.