Documentation One key, one bill, every major LLM provider. Everything you need to integrate Nemo Router into your application — auth, credits, observability, and guardrails.
Nemo Router is a managed LLM gateway that gives your team one API key to access 78+ models from Anthropic, Google, and OpenAI today — with AWS Bedrock shipping next and Azure OpenAI, Mistral, and Meta on the roadmap. New here? Start with Why Nemo Router Quick Start
Request flow What happens inside Nemo Router when your app calls a model.
Nemo Router request flow Your application sends a request to Nemo Backend, which authenticates the virtual key, reserves credits, applies guardrails and A/B routing, then routes to the chosen provider. The provider response returns with a cost header used to settle credits and log the call. Your application OpenAI SDK · sk-nemo-… POST /v1/chat/completions Nemo Backend (FastAPI) Auth · virtual key Reserve credits Guardrails A/B routing Prompt templates Provider routing forward Providers OpenAI Anthropic Vertex Bedrock Azure read response cost · settle credits · log · stream back Onboarding From sign-up to your first API call in five steps.
Nemo Router onboarding flow Six steps from signing up to making your first API call: sign up, verify email, choose plan, set up team, create your first virtual key, and make the first call. Sign up Supabase Auth 1 Verify email magic-link or pwd 2 Choose plan Tier 1 · 2 · 3 3 Set up team Default auto-created 4 Create key sk-nemo-… (once) 5 First call x-nemo-* headers 6 One org per user · Default team auto-created · key shown once Architecture The services in the request path and where data lives.
Nemo Router architecture The request path: your application connects over HTTPS to Nemo Backend (FastAPI on port 8090), which runs the provider-routing core in-process. Nemo Backend handles auth, guardrails, A/B, prompts, and credit reserve/settle against Supabase Postgres with row-level security. The routing core handles cost tracking, provider routing, fallback chains, and spend accounting before talking to OpenAI, Anthropic, Google Vertex, AWS Bedrock, or Azure OpenAI. Your application HTTPS · sk-nemo-… Nemo Backend · FastAPI :8090 Nemo concerns Auth · Guardrails · A/B · Prompts Credits (reserve + settle) Supabase nemo schema · RLS Provider router (in-process) Cost tracking · Provider routing Fallback chains · Spend Postgres public schema (Prisma) OpenAI · Anthropic · Google Vertex · AWS Bedrock · Azure OpenAI · … Request flow Onboarding Architecture
What happens inside Nemo Router when your app calls a model.
Nemo Router request flow Your application sends a request to Nemo Backend, which authenticates the virtual key, reserves credits, applies guardrails and A/B routing, then routes to the chosen provider. The provider response returns with a cost header used to settle credits and log the call. Your application OpenAI SDK · sk-nemo-… POST /v1/chat/completions Nemo Backend (FastAPI) Auth · virtual key Reserve credits Guardrails A/B routing Prompt templates Provider routing forward Providers OpenAI Anthropic Vertex Bedrock Azure read response cost · settle credits · log · stream back
Why Nemo Router Introduction Quick Start Authentication
Playground Guides API Reference SDKs
Need to leave the docs and look around the product? The footer below has every surface — or open one inline:
Model catalog Every model live in the router right now — context window, price per million tokens, provider, capabilities.
Open the catalog Was this page helpful? Yes No
