Skip to content
Router One
Sign up

Resell LLM API access: build your AI API business on a managed upstream

Resell LLM API access by building your own product on Router One, the managed multi-model upstream. You bring the portal, branding, pricing, and billing — your own one-api / new-api relay or your own app. Router One provides one OpenAI-compatible endpoint for 30+ models, multiple API keys with per-key budgets (maxSpend), and per-request usage data so you can bill your own margin on top. The pay-as-you-go token line uses posted model rates, with checkout-visible FX/channel fees kept separate; you set the customer margin.

Get your API key

What you bring vs what Router One provides

Router One is upstream infrastructure, not a white-label product. The line is intentionally clear so you keep full control of your customer relationship.

What you bring

  • Your own portal, app, or relay (one-api / new-api) — your branding
  • Your own customer pricing and the margin you charge
  • Your own billing, invoicing, and customer accounts
  • Your own signup, support, and terms with your customers

What Router One provides

  • One OpenAI-compatible endpoint for 30+ models
  • Multiple API keys per account with per-key budgets (maxSpend); rate limits follow platform defaults and can be raised on request
  • Per-request usage data (model, tokens, cost, latency) to drive your billing
  • Logs filterable by API key, with CSV export of request logs and the wallet ledger — reconcile each downstream customer from their key
  • Pay-as-you-go token line at posted model rates, with checkout-visible FX/channel fees kept separate

Why resell on Router One as upstream

CapabilityRouter OneOpaque relays
Real models, no swapping/downgradingThe model you request is the model servedOften silently downgraded or mixed
Per-request trace to self-verifyModel, tokens, cost, latency, status per requestLittle or no per-request visibility
Posted model ratesPosted-rate token line; you set your customer marginOpaque, bundled pricing
Per-key budgetsmaxSpend budget and expiry per keyCoarse or account-wide only
Published data handlingAPI bodies not retained; Playground history savedOften unclear
Models available30+ models (GPT, Claude, Gemini, Grok)Varies

Add Router One as a channel to your existing setup

Already running a relay (one-api / new-api) or your own app? Add Router One as one OpenAI-compatible channel — just a base URL and an API key. No SDK swap, no code rewrite. The base URL depends on who builds the request path: your own app or an OpenAI SDK uses https://api.router.one/v1. In one-api / new-api, a channel of type OpenAI takes the bare host https://api.router.one without a trailing /v1, because the relay appends /v1/chat/completions itself and a doubled /v1 turns into a 404; one-api's separate OpenAI-compatible channel type takes the /v1 URL instead. The API relay station guide (https://router.one/llm-api-relay-station) walks through the full channel setup, including a Claude-type channel for buyers on Claude Code and what Codex needs on the Responses path. Your existing pricing, accounts, and branding stay exactly as they are.

channel-config
# Base URL depends on who builds the request path
# one-api / new-api, channel type "OpenAI" (appends /v1/chat/completions)
Base URL:  https://api.router.one
# Your own app / OpenAI SDK (and one-api's "OpenAI 兼容" channel type)
Base URL:  https://api.router.one/v1
API Key:   sk-your-router-one-key

Per-key budgets do the metering

Create a separate API key per tier, customer segment, or environment. Each key carries its own maxSpend budget, so spend is capped before it runs away — and per-request usage data tells you exactly what to bill. Request and token rate limits follow platform defaults and can be raised on request. You set your customer pricing in your own portal; Router One just meters the upstream cost accurately.

Reselling vs the referral program

These are two different paths. The referral program pays 5% commission when someone tops up through your invite link — that is affiliate income from sharing a link. Reselling means running your own API business with Router One as the upstream infrastructure: your product, your pricing, your customers, your margin. If you just want to share a link, use the referral program.

See the referral program →

The transparent alternative to opaque relays

Real models, no swapping

The model you request is the model served — no silent downgrades, no mixing. Your customers get what they paid for, and you can prove it.

Request-data handling

Router One does not retain prompt or model-response bodies from direct API calls. It records request metadata for billing, usage tracking, and troubleshooting. Playground saves conversation history so users can revisit and continue their chats.

Per-request trace to self-verify

Every request carries a trace showing the model, tokens, cost, latency, and status. Verify the served model yourself instead of trusting a black box.

FAQ

Can I resell LLM API access through Router One?

Yes — by building your own product, app, or relay on top. Router One is the managed multi-model upstream: one OpenAI-compatible endpoint for 30+ models, multiple API keys with per-key budgets (maxSpend), and per-request usage data. You set your own pricing and branding. Router One does not provide white-label branding or sub-accounts — you bring the portal and the customer relationship.

How do I bill my own customers?

In your own portal or billing system, using the per-request usage data (model, tokens, cost, total time and, for streamed requests that produced output, time to first token) Router One exposes. Issue a separate API key per customer or tier; each key's maxSpend caps its spend so no single customer can run away with it. You charge whatever margin you choose on top of the posted upstream cost. Filter Dashboard → Logs by that customer's key and export CSV; Dashboard → Usage ranks spend by key.

How do I know when the upstream wallet is running low?

Poll GET https://api.router.one/v1/balance from your server. It returns the account's prepaid balance in USD as balance (spendable now), reserved_balance (held for in-flight requests) and total_balance, live and under its own limit of 60 requests per minute. Alert on balance and top up in the dashboard: there is no top-up API, and an empty wallet returns HTTP 402 on every customer key at once. Per-customer limits stay on each key's maxSpend; the balance monitoring guide (/blog/llm-api-balance-monitoring) covers thresholds and scripts.

Is there markup from Router One?

The pay-as-you-go token line uses posted model rates, with checkout-visible FX/channel fees kept separate. You decide the customer pricing in your own portal and keep the spread above your upstream cost.

What's the difference from the referral program?

The referral program is affiliate income: 5% commission when someone tops up via your invite link. Reselling is running your own API business with Router One as upstream infrastructure — your product, pricing, and customers. They are independent paths.

Do you store prompts or completions?

Router One does not retain prompt or model-response bodies from direct API calls. It records request metadata for billing, usage tracking, and troubleshooting. Playground saves conversation history so users can revisit and continue their chats.

Related

Build your AI API business on Router One

Get your API key