Pay-per-call data, web and file tools for AI agents. Each call is paid in USDC over HTTP 402 (x402), with no account or API key.

x402Base · Polygon · Arbitrumno signupMCP$1 file tools

Finance endpoints for trading agents

Endpoint	Price (USDC)	What it does
POST /v1/onchain	$0.01	Base wallet/token intel: native + ERC-20 balances, EOA/contract, tx count, token metadata.
POST /v1/dex	$0.01	Live DEX price / liquidity / 24h volume across all Base DEXs, with cross-DEX arbitrage spread + timestamps.
POST /v1/news	$0.02	Recent news headlines + heuristic bull/bear sentiment for a ticker or topic.
POST /v1/edgar	$0.02	Latest SEC filings (10-K/10-Q/8-K) for a US ticker, with document links.

Web and text tools

Endpoint	Price (USDC)	What it does
POST /v1/proxy	$0.002	URL → clean LLM-ready Markdown (renders JS).
POST /v1/proxy/stealth	$0.05	Cloudflare/anti-bot fetch via residential proxies + CAPTCHA solving → clean Markdown.
POST /v1/minify	$0.005 / 10k tok	Compress text before an LLM reads it. The response estimates the tokens it saved on your text.
POST /v1/dedup	$0.002 hit / $0.003 miss	Cache for repeated queries: an exact match, or near-identical wording. It hashes words rather than running a model, so the same question in other words is a miss.

$1 file tools

Fix a rejected upload, shrink a file, convert HEIC to JPG, remove a background, and more: /tools. Each is $1, by card for people or over x402 for agents.

LLM calls

Endpoint	Price (USDC)	What it does
POST /v1/infer	from $0.08 / 1M input tokens	Chat completions on Llama 3.1 8B and 70B, Qwen 2.5 7B and Mistral Small, relayed to OpenRouter (a hosted model API); gate402 runs no GPUs for this. The 402 quote prices your prompt plus max_tokens and you pay that quote; unused tokens are not refunded. Signed usage receipt. Rates per model: /v1.

Use it from Claude, Cursor, Cline …

Everything ships as an MCP server (20 tools). Drop it into any MCP client; a free-credit key is claimed automatically on first use, no wallet or signup needed:

{
  "mcpServers": {
    "gate402": { "command": "npx", "args": ["-y", "gate402-mcp"] }
  }
}

Finance: gate402_onchain · gate402_dex · gate402_token_risk · gate402_momentum · gate402_best_swap · gate402_launches · gate402_news · gate402_edgar.
Web tools: gate402_scrape · gate402_scrape_stealth · gate402_minify · gate402_dedup.
LLM calls: gate402_infer.
GPU marketplace: gate402_providers · gate402_become_provider · gate402_market_infer (see List your GPU).
Free (local, no key): gate402_token_count · gate402_html_to_md · gate402_json_repair.

The package also has gate402_compute, but /v1/compute answers 503 while no GPU worker is connected.

npm: gate402-mcp · source: github.com/pgalyen1987/gate402-mcp

Quickstart

Every call returns HTTP 402 with payment requirements. Pay with an x402 client, or present an X-API-Key.

1. See the price (any HTTP client):

curl -s -X POST https://gate402.app/v1/minify \
  -H 'content-type: application/json' \
  -d '{"text":"strip the filler from this verbose passage..."}'
# -> 402 { "x402Version": 2, "accepts": [ ...Base/Polygon/Arbitrum... ], ... }

2. Pay with x402 (e.g. x402-fetch or any agent x402 client), or with an API key:

curl -s -X POST https://gate402.app/v1/minify \
  -H 'content-type: application/json' \
  -H 'X-API-Key: sk_live_...' \
  -d '{"text":"..."}'

No wallet? Claim a free key with POST /v1/free-key and send it as X-API-Key. It holds $2.00 of credit for calls up to $1.00 each: two $1 file tools, or crypto and market data, SEC filings, news and web→Markdown. Stealth fetch, LLM calls and the five file tools that send text to an AI model need paid credit: add credit with a card or pay that call over x402.

List your GPU

The GPU marketplace is open for providers. No GPUs are online yet. Any provider that passes the check below is listed and routed automatically.

Buyers call POST /v1/market/infer and are charged only when a provider delivers. A call for a model that no listed provider serves gets 503, before any payment step. Container jobs (POST /v1/compute) answer 503: no GPU worker is connected.

You need	Details
An endpoint	OpenAI-compatible (vLLM, Ollama, TGI, LM Studio, llama.cpp) on a public HTTPS URL that answers POST /chat/completions for your model with no API key, since buyers send none, and GET /models with any status below 500 within 8 s. Its replies must carry the header X-Gate402-Wallet: <your payout wallet>, which shows the endpoint is yours, so nobody can list a server they do not run; set it in the reverse proxy in front of your model server (Caddy: header X-Gate402-Wallet 0x…). Registration checks all three; after that, listings are re-checked with a short test completion, and one that keeps failing comes off the list. Localhost, private addresses and redirects are refused. A tunnel or reverse proxy in front of a home machine is fine.
A payout wallet	An EVM address you control. It signs the registration on your machine (EIP-191) and receives 85% of each successful call in USDC on Base; gate402 keeps 15%. Owed balances go out in an hourly sweep once they reach $0.10.
A price	Per call, in USDC, at least $0.0134 (the fee has to cover payout gas). Buyers are routed by reliability first, then price, and a new listing's first calls count as a trial, so it can win on price from day one.

One command lists you. Use a wallet made for payouts: the key only signs the registration on your machine and is never sent.

npm i viem
curl -O https://gate402.app/register-provider.mjs
node register-provider.mjs \
  --endpoint https://your-host/v1 \
  --model llama-3.1-8b \
  --price 0.02 \
  --key 0xYOUR_PAYOUT_WALLET_PRIVATE_KEY

Change your price or endpoint, or delist, with POST /v1/providers/:id/update signed by the same wallet. Who is online: GET /v1/providers.

Independent trust scores

Reliability grades from x402.fuchss.app, an independent scorer: uptime, spec compliance and settlement record.

Machine-readable

Discovery manifest: /.well-known/x402.json
Full catalog + input schemas: /v1
For LLMs: /llms.txt