monkgatellm gatewayv1

Every open model.
One gate.

One key for four open-weight models, behind the OpenAI and Anthropic APIs you already use. Streamed as the model thinks, billed per token, paid from your wallet in USDG, or in $MONK for 25% more credit. No seats, no subscription, no email.

status
…
models
4
from
$0.14 / 1M
monkgate-omlive · no key needed

Ask the monk anything. A real model answers, token by token, and you see exactly what it would cost.

receipt appears here · 5 free questions per hour
01

Sign in. Top up. Call.

Three steps from a wallet to your first streamed token. No account form, no card, no subscription.

Your wallet is your account

  • Connect any wallet and sign one message. It is a signature, not a transaction: no gas, nothing moves.
  • The server checks the signature and a one-time nonce, then opens a session. Each nonce works once, so a signature can never be replayed.
  • No email, no password, no form. The address you sign with is the account that holds your balance and keys.

Diagrams illustrate the flow; they are not live data.

02

Pay with it. Hold it.

$MONK is the gateway's own utility: more credit when you top up with it, higher limits while you hold it. Launching soon.

01 Pay with itlaunching soon

+25%

credit on every top-up paid with $MONK

with USDG
$80.00
with $MONK
$100.00
02 Hold it

Keep $MONK in your signed-in wallet for 24 hours and your account moves up a tier.

tierholdreq / minper keyparallel
No token—300608
Holder1M60012016
Builder5M1,50030032
Scale20M3,00060064
03 Fair by design
  • Priced low, on purpose. A token deposit is valued at its lowest price of the last 30 minutes, so a brief spike can't buy extra credit.
  • Capped. Up to $100 of token credit per account per day, and $1,000 across all accounts. USDG has no daily limit.
  • Held, not flashed. Tiers look at what you held over 24 hours, so a quick buy-and-sell doesn't unlock anything.
Token details →

$MONK is a utility for paying for and using Monkgate. It is not an investment, and nothing on this site is a promise about its price or any return.

03

Endpoints, formats, billing, limits. The whole contract on one page.

modelforcontextin / 1Mout / 1Mrelative cost
Flagship for hard reasoning, coding and agents
Balanced quality and cost for everyday work
Capable and inexpensive general model
Fastest and cheapest, with a 1M-token context
POST /v1/chat/completions · POST /v1/messages · GET /v1/models
OpenAI and Anthropic wire formatsTool use, streaming and native error shapes. Your SDK never notices the switch.
Server-sent events, forwarded line by lineNothing is buffered. The first token leaves the moment it exists.
Per token, from a prepaid balanceWorst-case cost is held before the call and settled to the exact count after. Disconnect early, pay for what was generated.
USDG on Robinhood Chain · $MONK, launching soonTop up with $MONK for 25% extra credit, or USDG 1:1. Credited once the block is final, about ten minutes. Exactly once, never twice.
Higher limits for token holdersHold $MONK for 24 hours to raise your account to 600, 1,500, 3,000 requests per minute.
One wallet signatureNo email, no password. Each nonce works once.
SHA-256 hashed at rest, scoped, revocableShown once. Revoked keys fail on their very next request.
60 req/min per key · 8 concurrentPlus per-account caps, so one runaway script cannot drain you.
Prompts are never storedRouted only to zero-data-retention providers. Usage logs hold token counts, model and time.
04

One request, end to end.

A typical call, from key check to ledger. Ask the monk above and it replays with your own numbers.

○ sample requestt = 0 mstokens 0
first token
412 ms
total
1.65 s
tokens in → out
41 → 64
billed
$0.000024
  1. 01gateway

    Key hashed and looked up, rate limit checked, worst-case cost held.

  2. 02thinking

    The model reads your prompt. Nothing to stream yet.

  3. 03streaming

    Every token is forwarded the moment it exists.

  4. 04settle

    Exact count priced and written to the ledger. Hold released.

05

Point it here. Keep everything else.

Your SDK, your prompts, your code. Only the address, the key and the model name change.

app.py +2 −2
11import os
22from openai import OpenAI
33
4−client = OpenAI()
4+client = OpenAI(base_url="https://monkgate.ai/v1", api_key=os.environ["MONKGATE_API_KEY"])
55
66stream = client.chat.completions.create(
7− model="gpt-4o-mini",
7+ model="monkgate-zen",
88 messages=[{"role": "user", "content": "Hello!"}],
99 stream=True,
1010)

also works withOpenAI SDKAnthropic SDKClineAiderContinueLangChainLlamaIndexVercel AI SDK

06

Questions

10 / 10 matches

  1. 06.01

    What are the Monkgate models?

    Four curated open-weight models, each a step on the path: Nirvana for hard reasoning and agents, Zen for everyday work, Koan for low-cost general tasks, and Om when speed matters most.

  2. 06.02

    Which clients work with this?

    Anything that lets you set a custom OpenAI-compatible base URL (Cline, Aider, Continue, LangChain, the OpenAI SDKs) or an Anthropic base URL (the Anthropic SDKs). Set the model to one of the ids listed above.

  3. 06.03

    How am I billed?

    Per token, at the listed price, rounded up to one micro-dollar per request. Before a request runs, its maximum cost is held against your balance so parallel requests can't overspend. You only pay for tokens that were actually generated.

  4. 06.04

    How do deposits work?

    Send USDG on Robinhood Chain from the wallet you signed in with. You'll also need a little ETH on Robinhood Chain in that wallet to pay the network fee. The deposit is verified on-chain and credited once it is final, meaning its batch has been posted to Ethereum. That usually takes about 10 minutes. Don't send from an exchange: we match deposits to accounts by the sending address.

  5. 06.05

    How do I get USDG on Robinhood Chain?

    The easiest way is to bridge USDC from another chain (Ethereum, Base, Arbitrum and others) to Robinhood Chain with a bridge such as across.to; it arrives as USDG, 1:1 with the dollar. Bridge a small amount of ETH the same way to cover network fees.

  6. 06.06

    What is $MONK for?

    It is Monkgate's own utility token on Robinhood Chain, launching soon. Top up with it and you get 25% extra credit (valued at its lowest price of the last 30 minutes, up to $100 of credit per account per day). Hold it in your signed-in wallet for 24 hours and your account unlocks higher rate limits. It is not an investment, and nothing here is a promise about its price.

  7. 06.07

    What happens if I disconnect mid-stream?

    The model may keep generating on our side until it finishes, and you are billed for every token it produced, not only the ones you received. Set max_tokens to cap the cost of a request you might abandon.

  8. 06.08

    Why did I get an empty reply?

    Monkgate models think before they answer, and that thinking counts toward max_tokens. If max_tokens is very small (under about 200), the budget can run out before the answer starts. Raise max_tokens or leave it unset.

  9. 06.09

    Do you store my prompts?

    No. Only metadata is logged: token counts, the model and the time. Prompts are routed only to inference providers with a zero-data-retention policy, which process them in memory to answer and don't store them or train on them.

  10. 06.10

    Are there rate limits?

    Yes: 60 requests per minute per key by default (you can set a lower limit per key), 300 per minute per account, and 8 concurrent requests per key.