Managed LLM Gateway

A managed LLM gateway for routing, guardrails, and cost control.

nRouter gives engineering teams one OpenAI-compatible API for multi-provider model access, intelligent routing, in-path guardrails, hard budgets, and exact spend tracking.

Get started Read the docs

llm-gateway · request path

Managed production control plane

ClientOpenAI-compatible SDK

RoutingCost · latency · fallback

GovernanceBudget + guardrails

SettlementExact provider cost

One keyNo BYOKAuditable

API key: 1
Token markup: 0%
Provider clouds: 5
Controls: 24/7

The control layer

What is an LLM gateway?

An LLM gateway is the control plane between your application and model providers. It gives teams one authenticated API while centralizing provider selection, failover, safety checks, budgets, observability, and cost attribution. Instead of wiring every provider into every service, your application sends requests through one governed route.

LLM Gateway

Why production teams need an LLM API gateway

Direct provider integrations multiply credentials, SDK differences, retry logic, outages, billing systems, and compliance decisions. A managed gateway keeps those concerns in one control plane so teams can change models without rewriting application code.

OpenAI-compatible API

Built into the same managed gateway so teams can ship with one API key, clear controls, and an auditable request path.

Multi-provider access

Built into the same managed gateway so teams can ship with one API key, clear controls, and an auditable request path.

Centralized spend controls

Built into the same managed gateway so teams can ship with one API key, clear controls, and an auditable request path.

LLM Gateway

Routing, provider failover, and request architecture

Each request enters nRouter through an OpenAI-compatible endpoint, passes identity and policy checks, is routed by cost, latency, capability, or intent, and can fail over to an eligible provider when the primary path degrades. Usage is recorded after the response for accurate cost attribution.

Cost, latency, capability, and intent routing

Built into the same managed gateway so teams can ship with one API key, clear controls, and an auditable request path.

Circuit breakers and fallback chains

Built into the same managed gateway so teams can ship with one API key, clear controls, and an auditable request path.

Request-level logs and provider outcomes

Built into the same managed gateway so teams can ship with one API key, clear controls, and an auditable request path.

LLM Gateway

Guardrails, PII protection, budgets, and cost tracking

Apply controls before a request reaches a provider. nRouter can redact sensitive data, detect prompt injection patterns, enforce team and key budgets, and make blocked requests visible without losing the audit trail. Provider list pricing and token usage stay attributable to the request.

PII and secret redaction

Built into the same managed gateway so teams can ship with one API key, clear controls, and an auditable request path.

Prompt-injection and unsafe-request controls

Built into the same managed gateway so teams can ship with one API key, clear controls, and an auditable request path.

Hard budgets, usage ledgers, and spend analytics

Built into the same managed gateway so teams can ship with one API key, clear controls, and an auditable request path.

OpenAI-compatible API

Keep your OpenAI-compatible client

Move an existing integration by changing the base URL and API key. The request shape stays familiar while routing, guardrails, budgets, and provider failover run behind the gateway.

from openai import OpenAI

client = OpenAI(
    base_url="https://api.nrouter.ai/v1",
    api_key="NROUTER_API_KEY",
)

response = client.chat.completions.create(
    model="nrouter/auto",
    messages=[{"role": "user", "content": "Summarize this report."}],
)
print(response.choices[0].message.content)

Gateway comparison

Choose the control plane that fits your stack

Gateway	Best known for	nRouter perspective
nRouter	Managed routing, guardrails, budgets, and cost visibility	One OpenAI-compatible control plane with provider failover and governance in the request path.
OpenRouter	Provider and model discovery	Compare provider coverage, pricing, retention, and governance requirements before migrating.
Portkey	Governance, observability, and integrations	Evaluate managed routing, controls, and enterprise operating requirements together.
Helicone	LLM observability and request analytics	Pair observability depth with the routing, safety, and budget controls your gateway needs.
Vercel AI Gateway	Developer integrations and platform workflow	Compare portability, provider ownership, fallback behavior, and governance scope.

Pricing

Transparent pricing for the gateway layer

nRouter applies a flat 4% platform fee to pay-as-you-go credit purchases, added on top of provider usage. There is no per-token markup: inference is settled at the served provider’s published list price. See the pricing page for current plan terms and the cost calculator for worked examples.

FAQ

Common llm gateway questions

What is an LLM gateway?

An LLM gateway is the managed API layer between an application and model providers. It centralizes authentication, routing, safety policies, budgets, observability, and cost controls.

Does nRouter require provider API keys?

No. nRouter manages provider credentials. Your application uses one nRouter API key and an OpenAI-compatible endpoint.

Can I keep my existing OpenAI SDK?

Yes. Point the SDK at the nRouter base URL and replace the API key. The request and response format remains OpenAI-compatible.

How does nRouter handle provider outages?

Routing policies can use fallback chains and circuit-breaker behavior to move eligible requests to another provider or model when the primary path is unavailable or degraded.

How does nRouter protect sensitive data?

Guardrails can inspect and redact PII or secrets and detect unsafe request patterns before provider dispatch, with request outcomes available for audit and operations.

How is LLM gateway pricing calculated?

Pay-as-you-go credit purchases carry a flat 4% platform fee added on top. Model usage is settled at the served provider list price with no per-token markup.

Explore nRouter

Related platform pages

Smart routing

Route by cost, latency, intent, weights, and fallback policy.

AI guardrails

Redact PII and block unsafe requests before provider dispatch.

Enterprise controls

SSO, audit logs, tenancy, compliance, and procurement support.

Build on one gateway

Move from prototype to governed production AI.

Start with one OpenAI-compatible key, then add routing, budgets, guardrails, and observability as your workload grows.

Create an API key Compare plans