One key for every live model. Credit sales open at launch.
Network Agent

Pay $10.
Route $20.

Every top-up turns into twice the API credit, spendable on any live model. Put in $50 and route $100.

Read the API guide Holder rewardHold $NETWORKAGENT. Claim up to 5M tokens.Details ↗

$0

subscription or setup fees

credit on every top-up

Buy API credit

Pay as you go

Choose your credit

You pay

$20

You get

$40

or pay with crypto

Card · USDC · ETH · $NETWORKAGENT

Drops into your stack

An OpenAI-compatible endpoint. Keep your SDK, change the base URL and the key.

Pay the way you like

Card, USDC, ETH or $NETWORKAGENT from a connected wallet or a one-off address.

Prompts never kept

Billing needs token counts, not your words. Prompt and response text is never stored.

Stake · share · build

Stake $NETWORKAGENT.
Share 10% of top-ups.

Every confirmed credit purchase puts 10% into a reward pool. Stakers receive API credit in proportion to their verified share of the stake.

Share of every top-up

10%

Total staked

0

$NETWORKAGENT

Stakers

0

verified wallets

Reward pool

$0.00

in API credit

Connect a wallet to stake.

Staking opens with the token. Your wallet is how the pool recognises your share, so connect it now and nothing else is needed on the day.

Recent reward runs will be listed here after the first pool is distributed.

Hold · verify · claim

Hold $NETWORKAGENT.
Claim up to 5M tokens.

Keep a qualifying balance on Robinhood Chain for twelve hours straight, then sign one free message to claim promotional prompt tokens.

Hold at least

2M $NETWORKAGENT

Claim

500K prompt tokens

Hold at least

10M $NETWORKAGENT

Claim

5M prompt tokens

Connect before you verify.

The twelve-hour clock starts once the token is live and your balance qualifies. Claims open with it.

Continuous hold required12H

How a request travels

See the agent at work

One balance. Every model. Roughly half the spend.

networkagent.site/agent

Network Agent

More answers.
Less spend.

Ask for agent/auto and the agent sends the request to the model that fits it, then tells you which one answered.

REQUESTmodel: autoAGENTGPT-6 AstraQwen3.8 Max (090…Gemini 3.8 FlashClaude Fable 5.1Mistral Medium 3…

Workspaces

Early workspaces

Opens at launch

Workspaces that connect a wallet will appear here once credit sales open. Wallet handles only — never balances.

Eight routing profiles

Name the job.
Get the right model.

Auto, Balanced, Cheap, Fast, Best, Reasoning, Code or Vision. Each profile narrows the live catalogue before the agent ranks what is left.

AutoBalancedCheapFastBestReasoningCodeVision
Explore the profiles model: "agent/auto"

Watch one request

One prompt, routed
as it lands.

Ask anything. The endpoint, the key and the balance stay the same whichever model answers.

Your prompt

What does a routing agent actually do, in one sentence?

key na-live-••••1c7e · profile balanced

Reply streaming

You pay $10You route $20

One key. 53 providers.

OpenAI logoOpenAIQwen logoQwenGoogle logoGoogleAnthropic logoAnthropicMistral logoMistralZ.ai logoZ.aiDeepSeek logoDeepSeekNVIDIA logoNVIDIA

The catalogue

A model
for every idea.

Browse all 420 models
Xiaomi logo

MiMo-V2.6-Pro-UltraSpeed

Xiaomi

Live
ReasoningToolsVision

Input /M

$4.35

Output /M

$8.70

Context

1.0M

Xiaomi logo

MiMo-V2.6-Flash

Xiaomi

Live
ReasoningToolsVision

Input /M

$0.140

Output /M

$0.280

Context

1.0M

Xiaomi logo

MiMo-V2.6-Pro

Xiaomi

Live
ReasoningToolsVision

Input /M

$0.435

Output /M

$0.870

Context

1.0M

SpaceXAI logo

Grok 4.7

SpaceXAI

Live
ReasoningToolsVision

Input /M

$1.60

Output /M

$4.80

Context

500K

PrismML logo

Ternary Bonsai 2 27B

PrismML

Live
ReasoningToolsVision

Input /M

$0.075

Output /M

$0.500

Context

262K

Z.ai logo

GLM 5.3 FlashX

Z.ai

Live
ReasoningToolsVision

Input /M

$0.370

Output /M

$1.25

Context

1.0M

u

Pareto

unbiased

Live
ToolsVision

Input /M

$2.50

Output /M

$7.50

Context

262K

Inference.net logo

Schematron V2 Turbo

Inference.net

Live
JSON

Input /M

$0.030

Output /M

$0.150

Context

128K

Inference.net logo

Schematron V2 Small

Inference.net

Live
JSON

Input /M

$0.050

Output /M

$0.230

Context

128K

Newest live models, refreshed hourly. Prices are per million tokens and set by each provider.

From zero to first reply

Start building
in four steps.

No subscription and no new SDK. Bring an OpenAI-compatible client and pick how you pay.

01

Connect a wallet

One click on Robinhood Chain opens your workspace. No email, no password.

02

Pick how to pay

Card, USDC, ETH or $NETWORKAGENT. Every top-up is doubled into credit.

03

Create one key

Copy it once and keep it on your server. Revoke and rotate whenever you like.

04

Call any live model

Same endpoint, same key. Change the model field, or let the agent choose.

Test before you ship

Bring a key.
Make it move.

Try the key in a sandbox with a real model, or put it to work in a head-to-head.

Catalogue

Connected

420

models one key can reach today

Providers53

Routing profiles8

Recent top-ups

Opens at launch

Card and on-chain credit purchases, newest first.

No purchases yet — checkout has not opened.

Small change, every model

Two lines.
You’re connected.

Point an OpenAI-compatible client at Network Agent. Your code keeps running; the models behind it multiply.

Read the API guide
.env
# Point any OpenAI-compatible client here
OPENAI_BASE_URL=https://api.networkagent.site/v1
OPENAI_API_KEY=na-live-••••••••

Before you ask

Simple on purpose.

What exactly am I paying for?+

Prepaid API credit. It is spent per token at each model's listed rate, and every top-up is doubled at checkout. There is no subscription and nothing renews on its own.

When can I buy credit?+

Checkout opens at launch. Wallet connection already works, so your workspace is ready the moment it does.

How does staking work?+

Staked $NETWORKAGENT shares a pool made from 10% of every confirmed top-up. Your share of the pool follows your share of the verified stake.

What are holder rewards?+

Holding $NETWORKAGENT for twelve hours unlocks promotional prompt tokens, up to five million per wallet. Claiming needs one signed message, never a transaction.

Do you keep my prompts?+

No. We record usage metadata for billing — model, token counts, timing — and never the text of a prompt or a response.

How does the agent choose a model?+

It filters the live catalogue by what the request needs (context length, tools, images, JSON), then ranks what is left by the profile you asked for. The model it picked is named in the response.

Does one key reach every model?+

Yes. The key is tied to your workspace balance, not to a provider. Any live model in the catalogue answers to it.