gate402 0.1.0

Python client for Gate402 — pay-per-call GPU/CPU compute, open-model LLM inference, and crypto/finance data for AI agents over x402 (USDC on Base). No signup, free tier.

gate402

Python client for Gate402 — pay-per-call GPU/CPU compute + open-model LLM inference, plus crypto/finance data and web tools, for AI agents over x402 (USDC on Base). No signup: a free-tier key is auto-claimed on first use.

pip install gate402

Quick start

from gate402 import Gate402

g = Gate402()  # free-tier key auto-claimed on first call (or set GATE402_API_KEY)

# Open-model LLM inference, billed per token
print(g.infer(prompt="Say hello in five words.")["text"])

# Rent GPU/CPU compute for a container job, billed per second
job = g.compute(image="ghcr.io/acme/vllm:latest", gpu="rtx4090", duration_sec=300, model="llama-3.1-8b")

# On-chain wallet / token intel on Base
print(g.onchain("0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913"))

Methods

Method	Endpoint	What it does
infer(prompt=…, model=…)	/v1/infer	Per-token LLM inference (Llama 3.1 / Qwen / Mistral)
compute(image, gpu, duration_sec)	/v1/compute	Rent metered GPU/CPU per second; signed receipt
onchain(address)	/v1/onchain	Wallet/token intel on Base
dex(address)	/v1/dex	Live DEX price / liquidity / volume
token_risk(address)	/v1/token-risk	Rug/tradeability verdict + honeypot sim
momentum(address)	/v1/momentum	Momentum + order-flow signal
best_swap(address, size_usd=…)	/v1/best-swap	Best pool + price impact for a trade size
launches()	/v1/launches	Freshest Base token launches
news(query) / edgar(ticker=…)	/v1/news, /v1/edgar	News+sentiment / SEC filings
scrape(url, stealth=False)	/v1/proxy	Fetch a page as clean Markdown
minify(text) / dedup(query)	/v1/minify, /v1/dedup	Token compression / semantic cache
catalog()	/v1	Free. Machine-readable endpoint catalog
providers()	/v1/providers	Free. Browse the GPU marketplace supply
market_infer(model, prompt=…)	/v1/market/infer	Marketplace inference — x402-native (see below)

Billing

The payment is the auth — there are no accounts to create. On first use the client self-claims a free-credit key from Gate402 and reuses it. When credit runs out, calls raise Gate402PaymentRequired with a top-up link. Pass api_key=… or set GATE402_API_KEY to use a funded account.

from gate402 import Gate402, Gate402PaymentRequired
try:
    g.infer(prompt="…")
except Gate402PaymentRequired as e:
    print(e)  # → top-up link

GPU marketplace

providers() lists third-party GPU/compute supply (free). The buyer endpoint market_infer() is x402-native — it settles on-chain per call with escrow-on-success, so the free-tier / API-key rail cannot pay it and it raises Gate402PaymentRequired. Use an x402 payment client to call POST /v1/market/infer directly. To sell your GPU, register at POST /v1/providers/register.

Configuration

Env var	Default	Purpose
GATE402_API_KEY	(auto free tier)	A funded account key to skip the free tier
GATE402_BASE_URL	https://gate402.app	Override the gateway base URL

MIT licensed.