Why nRouter
One API key, every model, control in the request path.
Route across models, enforce pre-flight budgets, and apply inline guardrails through one managed gateway with a 0% platform fee.
Five problems. One control plane.
The five failure modes that break production AI and autonomous agents, and how nRouter closes each one.
Spend is governed before the call.
AI is metered by the token. Pre-flight budgets enforce hard spending caps per org, team, and key.
- Reserve credits before the call and settle at exact list price
- Calls over budget return HTTP 402 with $0 held and $0 spent
- Organization, team, and virtual-key budget scopes
Every model behind one API key.
One unified nRouter key gates the live catalog with no BYOK. Switching models is a single string.
- Keep the OpenAI SDK you already use and point one base URL
- Provider credentials stay managed inside nRouter
- Scope allowed models and routes per virtual key
Guardrails run inline on every request.
Built-in PII redaction, prompt-injection defense, and content moderation run in the request path.
- Pre-call checks block or redact before a provider is called
- Requests blocked before credit reservation hold and spend $0
- Configure rules across key, team, and organization scopes
Failover and routing at one seam.
Choose a model with configured fallbacks, or send nrouter/auto for task fit or lowest list rate.
- Pre-stream provider faults advance to your next configured fallback
- Attempts we fail over from before the first byte cost you $0
- Up to 4 fallback models per organization or per request
One console. Real list price attributed.
Exact model list price returned in response headers and attributed to every key, team, and org.
- 0% platform fee on credits, from a $5 minimum purchase
- Response headers report served model, tokens, and settled cost
- Request logs record operational metadata without prompt text
Whoever asks why, the answer holds up.
One key, one bill, and one bar of control over every model your team touches.
The team adopting it
One key replaces separate provider accounts and invoices. Hard budgets, guardrails, and RBAC ship built in.Will this simplify our stack?The systems architect
Rust data plane with pre-flight access checks, inline guardrails, and configured fallback chains.How does the request path run?Finance & procurement
0% platform fee on credits and exact model list prices, with per-key and per-team spend attribution.How is usage billed?The builder shipping code
Point your existing OpenAI SDK at our base URL and call every live model. Pay as you go from $5.How do we integrate?
Built for autonomous workloads.
AI is becoming autonomous; control in the request path must be automatic.
Zero platform markup.
Pay as you go is the one self-serve plan. Credits purchase at face value and settle at each served model’s published list price.
0%
Pay as you go.
Start without subscriptions or per-seat charges. Every governance control—routing, guardrails, budgets, SSO, and audit logs—is included.
$5
Pre-flight protection.
Key access, rate limits, budgets, and configured input guardrails run before a provider is called.
$0