About Router One

Router One is a unified LLM API gateway: one OpenAI-compatible endpoint (https://api.router.one/v1) for GPT, Claude, Gemini, Grok and the other families in the live catalog, with same-model fallback, per-request cost traces and per-key spend caps. It serves developers and startup teams building on LLMs, and resellers who run their own AI API service on a managed upstream. Pay per token from a prepaid wallet — up to 90% off official prices on select models — or on a Pro, Max or Ultra plan; the gateway is reachable worldwide, including mainland China.

What Is Router One?

Router One provides a single API endpoint that connects to the model families in the live catalog — GPT, Claude, Gemini, Grok and DeepSeek among them. Instead of managing separate SDKs and dealing with provider-specific authentication, rate limits, and billing, your team uses one consistent interface. The gateway is reachable worldwide — including mainland China — and takes payment by card, Alipay, or stablecoins, so neither geography nor billing blocks a team from frontier AI models.

Core Capabilities

Our platform is built around three pillars: Exact-model routing that serves each request with the model it names and fails over between healthy routes for that model; Spend Control, where every API key carries a maxSpend cap you set on its cumulative wallet charges (plan-covered requests do not count) while rateLimit and tokenLimitTpm run at platform defaults that can be raised on request; and Full Observability with request traces, aggregated metrics, and cost breakdowns. Create a separate key for each project, tool, or environment to isolate its spend. We also provide native support for AI coding tools — Claude Code and Codex connect with a one-line install script or a few lines of configuration.

Our Philosophy

Router One gives teams one place to review model-call usage and control spending. Customer Logs shows the model, request ID, tokens, wallet charge or plan quota used, recorded latency, status, and any sanitized error. Upstream supplier identities and service routes remain internal. We believe developers deserve clear evidence for the usage and charges behind their AI applications.

What's Next

We focus on making the gateway better at the job it already does: tighter latency, broader provider coverage, sharper observability, lower integration friction. We do not promise agent-orchestration runtimes, Run/Step state machines, or platform layers we are not building. When we ship something, it is because we ship it — not because we put it on a roadmap slide.

Who writes this site

Router One's pages are written and maintained by the Router One engineering and documentation team — the same people who run the gateway. There are no named guest authors. Every published latency, pricing, routing, and retention number traces back to a dated page: the China latency benchmark, pricing methodology, smart routing methodology, and data retention windows are indexed under Methodology and the Trust hub. Corrections and questions go to support@router.one.

Contact & official accounts

Contact Email

support@router.one

X (Twitter)

x.com/router_one

GitHub

github.com/Router-One