Built for Elysium mainnet · chain 99901
The first LLM router
on Elysium
One key for every model. Pick a route and Userouter sends each prompt to the model that fits, paid in HYPE and settled onchain.
Featured
- APIBuilding
Router API
One OpenAI-compatible key for every model, with usage paid for on Elysium mainnet.
Learn more → - PaymentsBuilding
HYPE top-ups
Send HYPE from your wallet on Elysium and your inference balance goes up by its USD value.
Learn more → - AgentsNext
Agent launchpad
Agents launch a token paired with HYPE, and its creator fees pay for their own inference.
Learn more →
Routing modes
Three routes. One key.
Tell Userouter what matters for this request and it picks the model. Same prompt, same code, a different trade-off per call.
Prompt
Refactor this TypeScript function and fix the bug
Cheapest
Claude Haiku 5.5
$0.20 / 1M
Lowest price that fits the task
Balanced
DeepSeek V4 Pro
$0.87 / 1M
Middle of the pack on price and depth
Smartest
GPT-6 Astra
$20.00 / 1M
The frontier model for the task
Models
Every frontier model, one key.
6 featured here, 435 in the catalogue.
openai
GPT-6 Astra
codingagentsresearch- Input / 1M
- $10.00
- Output / 1M
- $50.00
google
Gemini 3.8 Flash
visionlong-contextfast- Input / 1M
- $0.75
- Output / 1M
- $3.75
x-ai
Grok 4.6
researchdata- Input / 1M
- $2.00
- Output / 1M
- $6.00
deepseek
DeepSeek V4 Pro
codingfastdata- Input / 1M
- $0.50
- Output / 1M
- $1.98
moonshotai
Kimi K3
agentslong-contextcoding- Input / 1M
- $3.00
- Output / 1M
- $15.00
anthropic
Claude Haiku 5.5
fastcodingagents- Input / 1M
- $0.10
- Output / 1M
- $0.50
Quickstart
Switch in two lines.
Point any OpenAI-compatible client at Userouter: change the base URL and the key, keep the rest of your code.
# before — OpenRouterOPENAI_BASE_URL=https://openrouter.ai/api/v1OPENAI_API_KEY=sk-or-v1-… # after — Userouter. Same two variables: your OpenAI-compatible client reads these.OPENAI_BASE_URL=https://api.userouter.xyz/api/v1OPENAI_API_KEY=sk-userouter-…
The stack
Four layers under every request.
From the route you pick to the chain it settles on. Each layer does one job, and you can check every one.
- 01RouterReads each request and picks the model for the route you asked for: cheapest, balanced or smartest.→
- 02GatewayA thin, OpenAI-compatible relay. Your request reaches the model you were routed to unchanged, and only billing metadata is kept.→
- 03CREDITTokenized inference: one CREDIT is one dollar of usage, earned by $ROUTER stakers and spendable through any key.→
- 04ElysiumThe Layer 2 that settles to HyperEVM, with HYPE as gas. Balances, stakes and payments settle here.→
FAQ
Questions, answered.
What is Userouter?
An LLM router on Elysium: one API key for every model in the catalogue, with each request sent to the model that fits the route you choose.
How does routing decide?
It reads what the prompt is for, such as code, long documents or quick chat, then picks the cheapest, the middle or the frontier model for that use case. The home page box shows the choice before you send anything.
Can I still pick a model myself?
Yes. Send a model name as you do today and the request goes to that model, unchanged. Routing modes are an option, not a requirement.
What do I change in my code?
The base URL and the key. Model names, streaming, tool calls and vision work as they do with any OpenAI-compatible client.
How do I pay?
With a balance on your account, topped up from your wallet on Elysium. Gas on Elysium is HYPE.
Do you store my prompts?
No. The gateway keeps billing metadata only: the model, token counts, cost and timing. Prompts and responses pass through and are not written anywhere.
Is it live?
Userouter is being built for Elysium mainnet. The roadmap shows what is live, what is in progress and what comes next, with dates only once something ships.
