Every model,
refracted through
one endpoint.
Point your agents at Chroma402. Many models behind an OpenAI-compatible API, metered by the token. No signup, no subscription, no card.
Pay for tokens, not for access.
No plans, no credit packs, no seats. Every call is billed for the tokens it actually used, at the price on the tag.
Send, price, sign, answer
Send a request, get a price, sign it, get your answer. No key, no account, no gas.
Deposit once, use anywhere
Deposit once and get a chroma_sk_ key that works in any OpenAI SDK. Every call spends exactly its token cost.
Repeated calls cost less
Byte-identical deterministic calls are answered from cache at a lower price, because agent loops repeat themselves more than you'd think.
Unused credit returns
Pay-per-call signs a ceiling. What the call didn't use is credited back to your wallet, instead of being kept.
$CHROMA: hold it, pay less.
The router token gives qualifying holders a discount on calls, read from the chain at call time. Inference itself is always paid in the settlement currency — you never need the token just to use the router.
Discount: 20%
Qualifying holding: 1,000,000 $CHROMA in the wallet that owns your key.
Contract address: TBA
Read the contract before the chart.
Launchpads mint thousands of tokens every day. Scan reads a contract straight from the chain — supply, owner, upgrade hooks, mint, blacklist and fee switches, then pool depth and recent flow — and summarises what it found in three sentences.
Enter an address to see a scan summary.
Fresh in the spectrum
Prices in USDG per 1M tokens. What's posted is what you pay.
A large open model with tool use and vision support, built for agent work.
A flagship agent coding model with a very long context window.
DeepSeek's strongest open model in this catalog.
A 120B mixture-of-experts model with 12B active parameters, built for agents.
A fast, small open model for everyday agent tasks.
A larger open model for harder reasoning and coding.
A strong multilingual open model for agents.
A balanced open model with multimodal support.
Your SDK already speaks Chroma402.
Bring a wallet, not an email
There are no accounts. The wallet that pays is the owner, and the key is the account.
Deposit, mint a key
Fund a prepaid chroma_sk_ key. It is shown once.
Swap one URL
Point any OpenAI client at Chroma402. Or skip the key and let your agent pay each call with x402.
from openai import OpenAI
client = OpenAI(
base_url="https://chroma402.pages.dev/v1",
api_key="chroma_sk_your_key",
)
reply = client.chat.completions.create(
model="chroma/llama-3.1-8b",
messages=[{"role": "user", "content": "What is Chroma402?"}],
)
print(reply.choices[0].message.content)
Billing is exact: tokens used × posted rate. No per-call fee, nothing recurring, nothing hidden.