Most AI proxy providers obscure real token costs through opaque token markups, bundled credit bundles, or feature-gating essential enterprise governance behind expensive monthly platform licenses. Transparent AI pricing demands a cost-honest model: charging exact upstream provider list prices for model tokens and taking a modest, fully disclosed platform fee to operate high-availability routing infrastructure.
Many gateway vendors silently add a 10% to 30% margin on top of standard provider token prices, turning high-volume production scaling into an unsustainable expense for fast-growing companies.
Withholding basic security essentials—such as audit logs, SSO, or budget caps—behind $2,000/month Enterprise tiers forces growing startups into insecure architectures.
Managing separate credit lines, payment cards, and invoicing terms across OpenAI, Anthropic, Google Cloud, and AWS creates enormous administrative overhead for accounting teams.
Prepaid credit systems that expire unused balances after 30 days artificially inflate effective token costs and penalize fluctuating application usage patterns.
Token spend is settled at the exact published upstream rate (e.g. GPT-4o, Claude 3.5 Sonnet, Gemini 2.0).
Transparent 4% fee charged on top of prepaid credits—never deducted from token balances or hidden in markups.
Every account tier receives identical access to all guardrails, smart routers, virtual keys, and audit logs.
nRouter operates with total pricing transparency: 100% exact provider list prices with zero per-token markup. All self-serve plans pay a flat 4% platform fee added to credit purchases, with every governance, routing, and observability capability unlocked from day one. You pay only for the exact tokens your models consume, billed at published provider rates, without hidden markups or paywalled safety features. With Pay as you go starting at a simple $5 credit purchase, teams can scale from prototype to enterprise with predictable unit economics.
Learn how our zero-markup pricing model compares to legacy gateways and calculate your team's monthly cost savings.
Prepaid tokens are one provider's currency, priced against one rate card. Gateway credits are dollars that fund any model behind one key. Here is how to consolidate, read the balance card, and prove each call settled at real cost.
Guardrails, audit trails, budgets, evals, A/B tests and prompt management are on every nRouter plan. Plans change the monthly allowance and the rate limits, never the feature set — here is the reasoning, the fee math, and what the bet costs us.
A gateway can take its cut two ways: silently, inside the per-token rate you can never decompose, or visibly, as a platform fee added at purchase. nRouter does the second, so every credit you buy is spendable at the provider's own settled cost.
If you resell AI, the model bill is the wrong unit. Attribute every call to the customer it served with the OpenAI user field, cap each customer independently, and reconcile the sum against your ledger to the cent.
Head-to-head: nRouter vs Cloudflare AI Gateway, on the axes that actually differ — an account-scoped endpoint, the cf-aig-* cache, Unified Billing's 5% credit fee and its 200-requests-per-60-seconds cap, and a per-account log pool. One base-URL switch, governance on every plan.
Head-to-head: nRouter vs Vercel AI Gateway on the axes that decide it — BYOK spend that budgets cannot cap, governance gated to Pro and Enterprise, OIDC auth that assumes a Vercel deployment, and what a zero-markup gateway costs you elsewhere. One base-URL switch.
Head-to-head: nRouter vs Helicone on the axes an observability-first product makes you choose — guardrails and evals starting at $79/mo, a one-month retention window on Pro, ingestion and API ceilings, and one seat on Hobby. Sourced against Helicone's own pricing page.
Every nRouter plan pays the same 4% platform fee on credits, so a subscription never lowers the fee. What Starter, Pro and Max buy is a monthly nrouter/auto usage allowance and higher rate limits — worth it only for traffic you are happy to let nRouter route.
A rebuilt savings model for a $40k/month LLM bill. Batch and prompt caching carry published rates; provider capacity reservations do not, so that lever is a labelled assumption with a stated range. Every line is separated into sourced and assumed.
Head-to-head comparison: nRouter vs OpenRouter, Portkey, Helicone. Guardrails, A/B tests, prompt management, evals, budgets — included on every plan. 4% of your credits on every plan, no minimum fee. One base-URL switch.
The engineering case for nRouter: routing is solved, but tenant isolation, non-negative credits, guardrails, and unified billing across providers are not.
