How do I get an API key?

In the dashboard: API Keys → Generate Key. The full key is shown exactly once — copy it immediately. Keys begin with sk-nrouter-, and every request carries one in the Authorization header. See Authentication.

Which SDK should I use?

The published TypeScript (@nrouter_ai/sdk) or Python (nrouter-sdk) SDK — or the OpenAI client you already have, with only the base URL and key changed. Other languages have preview SDKs and a cURL reference; see the SDK index.

Why should I use nRouter?

One key reaches models across OpenAI, Anthropic, AWS Bedrock, Google Vertex AI, Azure Foundry, and Alibaba US. Every call settles at the exact list price with 0% markup, and budgets, guardrails, fallbacks, and request logs are built in rather than bolted on. See Why nRouter.

Which models can I use?

Whatever the live catalog shows today is what you can call — browse it on the Models page or fetch it from the public models API. Send a model ID from the catalog, or nrouter/auto and let the router choose.

What does nrouter/auto do?

The router picks the model for each request. The model that actually served is named in the x-nr-model response header and in your request logs — nothing is hidden. To fix the model yourself, send a catalog model ID instead.

How do fallbacks work?

List models in the nrouter_fallbacks request field, in order. On a failover-eligible error — an upstream 429, 503, or Anthropic 529, or a connect failure — the request moves to the next entry. A request makes at most 3 provider calls in total. See Per-Request Options.

How does streaming work?

Set stream: true on a chat request and tokens arrive as server-sent events (SSE) as the model produces them. The request is logged like any other, with the model that served it and its exact cost. See Chat Completions.

How do rate limits work?

Limits are enforced per key (requests and tokens per minute), and per team, user, and budget. A 429 tells you which ceiling refused in x-nr-limit-source and how long to wait in Retry-After. See Errors & Rate Limits.

What formats can I send in a request?

Text, always. Images too, for the catalog's vision models: in the message content, as base64 data or a hosted URL. See Send Image Input.

What does nRouter charge for models?

The provider's exact list price. Token markup is 0% and the platform fee is 0% — the price shown for a model is the amount deducted from your balance.

Pay as you go. You add credits, and each request deducts its exact cost as it completes. Every request's cost is recorded in your request logs and in the x-nr-request-cost response header. See Billing.

Are there plans or tiers?

One public plan: pay as you go. There are no feature tiers — budgets, guardrails, fallbacks, and analytics work the same for every account.

Can I use my own provider keys (BYOK)?

No — nRouter does not do BYOK. One nRouter key reaches every provider cloud, which is what lets budgets, logs, and guardrails apply uniformly to all your traffic. We published the reasoning in Why nRouter does not do BYOK.

Why did my request return 402?

The error code says which: insufficient_credits means the account is out of credits; budget_exceeded means a budget refused the spend. The x-nr-limit-source header names the ceiling that refused. See Errors & Rate Limits.

Where do I see what a request cost?

In the dashboard's request logs, and in the x-nr-request-cost header on the response itself. Every call settles at the exact list price, with 0% markup.

What data is logged about my requests?

Request logs record the metadata: the model, token counts, exact cost, and the key alias used. What content is retained is controlled in Settings under Privacy & data policy, and administrators can export data for GDPR requests.

What analytics are available?

The Advanced section turns your request history into reports — Cost, Usage, Performance, Key Usage, Budgets, Benchmark, and Compare, plus the API Consumers report for per-end-user attribution. Every report shares one filter bar (time window; group by model, provider, key, team, or tag) and exports to CSV. See Analytics & Reports.

How does team access work?

Four roles — Viewer, Member, Org Admin, Owner — decide who can see and change what, and keys can be scoped to a team. See Team Management.

How do I delete my organization?

In Settings, open the Danger zone (/settings/danger-zone) and choose Delete organization. It is irreversible: the organization must have a single member — remove everyone else first — and you confirm by typing the organization slug. After deletion you are returned to the sign-in page.

Docs FAQ

Answers to the questions nRouter users ask first: getting a key, models, streaming, fallbacks, pricing with 0% markup, rate limits, and request costs.

New to nRouter? These are the questions users ask first, grouped by topic — each answered in a few lines, with a link to the page that covers it in full.

Quick Start — Make your first API call
Authentication — How API keys and auth work
Errors & Rate Limits — Every status, code, and retry rule

Docs FAQ

Getting started

Models

API

Pricing & billing

Privacy & account

Next steps

On this page

Docs FAQ

Getting started

How do I get an API key?

Which SDK should I use?

Why should I use nRouter?

Models

Which models can I use?

API

What does nrouter/auto do?

How do fallbacks work?

How does streaming work?

How do rate limits work?

What formats can I send in a request?

Pricing & billing

What does nRouter charge for models?

How am I charged?

Are there plans or tiers?

Can I use my own provider keys (BYOK)?

Why did my request return 402?

Where do I see what a request cost?

Privacy & account

What data is logged about my requests?

What analytics are available?

How does team access work?

How do I get support?

How do I delete my organization?

Next steps

On this page