Your first $5 becomes $15Get started
Playground

Test any model. Right from your browser.

An interactive playground for 240+ models. Chat, tweak parameters, compare responses, watch cost + latency in real time, and export integration code in 15 languages — without writing a single line of integration code.

playground · gemini-2.5-flash

Active session

Auth modesk-nemo-...x4kf
Key storagesessionStorage
Streamingon (SSE)
Tokens streamed142
Cost so far$0.0021
Latency (first token)420 ms
Code generators15 languages
streamingnon-streamingcomparecode-gencancel
Server-side key storage
0

Your key lives in sessionStorage only

Models you can test
240+

Alibaba US, OpenAI, AWS Bedrock, Azure Foundry, Google Vertex AI & Anthropic

Streaming + non-stream
Both

Real-time tokens or full-response mode

Code-gen languages
15

8 SDKs + 7 frameworks

The session

Chat, swap models, watch the cost — per message.

playground · session 7f2agemini-2.5-flash ▾
Summarize this support ticket and suggest a reply tone.

The customer reports a billing mismatch after upgrading mid-cycle. Suggested tone: empathetic and direct. Confirm the proration math first, then offer the credit…

218 tokens · $0.0008 · TTFT 120 ms

Now try it on gemini-2.5-pro and compare.
streaming on gemini-2.5-pro…
temperature 0.7max_tokens 1024session cost $0.0021export code → 15 languages
Request pipeline

The same path as production: no sandboxing

Every playground request uses your real virtual key and traverses the same auth, credit-reserve, guardrail, and provider chain as your SDK. No fake responses, no mock providers, no silent shortcuts.

Playground request flow

  1. Paste key

    sessionStorage only

    Cleared on tab close; never sent to nRouter servers.

  2. Frontend (Next.js)

    Cloud Run

    Forwards request with the user-pasted virtual key.

  3. Nemo Backend

    nRouter

    Auth, credit reserve, guardrails: same as production.

  4. Nemo routing core

    Routing engine

    Cost & usage tracking, rate limits, provider fallback.

  5. Provider

    Vertex / Anthropic / OpenAI

    Real LLM call. No sandboxing, no fake responses.

Security model

Your key never leaves the browser

Virtual key only

sessionStorage scope, master key isolated

Every LLM call (playground, SDK, or production API) must authenticate with a virtual key (sk-nemo-...). The playground enforces this by holding the key only in browser sessionStorage. Refresh the tab and you re-paste; close the tab and the key is gone.

  • Browser sessionStorage scope (cleared on tab close)
  • Never POSTed to nRouter for persistent storage
  • Master key never enters the playground flow
  • Per-key spend, RPM, TPM enforced exactly as in production
security · session

Storage + auth posture

Key in sessionStorageyes
Key in localStoragenever
Key in cookiesnever
Server-side persistence0 paths
Master key in this flowisolated
Cleared on tab closealways
Virtual keys onlysk-nemo-... onlymaster key isolated
Full playground reference: chat, streaming, parameters, code-gen
Full chat interface
System prompts, multi-turn history, model swap mid-conversation, and a side-by-side compare model. Latest-response tokens, cost, and latency in the Metrics tab. Cancel mid-stream to stop token spend.
Streaming + non-streaming
SSE streaming (default) with a live token-rate counter, or full-response mode for batch / eval workflows. Identical billing. Both modes use the same pipeline.
Parameter pass-through
temperature, top_p, max_tokens, and a system prompt. Every parameter you set is forwarded as-is to the provider, with no silent defaults.
Code generation — 15 languages
Python (native + OpenAI-compatible), Node.js, Go, Ruby, Java, PHP, cURL, plus 7 frameworks (LangChain, LlamaIndex, CrewAI, AutoGen, Google ADK, more). Drawn from the canonical SDK examples package: embeds your exact prompt + parameter set.
Error surfacing
402 (out of credits) surfaced with an inline top-up CTA. 429 with a retry-after timer. Provider 4xx errors shown verbatim. Streaming aborts cleanly on mid-stream errors.
Model catalog access
Every model in your account, one dropdown. Filter by capability (chat, vision, embedding, image, video), provider, latency budget, or cost ceiling. Switch models mid-conversation; price-per-1k tokens visible per model.
FAQ

Common playground questions

Is the playground free?

Yes. There is no additional fee for using the playground itself. You pay only for the tokens consumed at the same per-token rate as API calls. Cost and latency are displayed in real time as you test, so there are no surprises.

What models can I test?

All 240+ models live on nRouter today across Anthropic (Claude), Google Vertex AI (Gemini, Imagen, Veo, embeddings), and OpenAI (GPT, o-series). Every model available through the nRouter API is available in the playground. AWS Bedrock (Claude, Llama, DeepSeek, Nova, Qwen) is live too.

Where is my API key stored?

Browser sessionStorage only. The key is cleared on tab close, never sent to nRouter servers for storage, and never persisted across sessions. This is a critical security requirement: all LLM calls must use virtual keys (not master keys), and the playground deliberately holds zero server-side state.

Does the playground support streaming?

Yes. Server-sent-events streaming is the default. You see tokens appear as the model generates them, with a live token-rate counter. Cancel mid-stream to save on cost. A non-streaming mode is also available for batch and eval workflows.

Can I export code from the playground?

Yes. Auto-generated integration code in 15 languages: Python (native + OpenAI-compatible), Node.js, Go, PHP, Java, Ruby, cURL, plus 7 frameworks including LangChain, LlamaIndex, CrewAI, AutoGen, and Google ADK. The generated code uses your exact prompt, model selection, and parameter configuration, ready to paste into your project.

240+ models · 15 code generators · 0 server-side keys

From idea to working prototype in under a minute

No SDK to install, no provider account to manage. Sign up, paste your virtual key, start chatting.