A shield for your agent before it reads the open web.

Prompt-injection scanning and clean webpage fetching, purpose-built for AI agents that have to read untrusted content. Screen it, fetch it clean, or both in a single call.

$ read the docs llms.txt for agents

trustfetch.duckdns.org

$ curl -s https://trustfetch.duckdns.org/tools/safe-fetch \

-X POST -d '{"url":"https://example.com"}'

HTTP/1.1 402 Payment Required

accept: 0.03 USDC on eip155:8453 -> 0x1c4A…e29

# client settles, retries with proof of payment

HTTP/1.1 200 OK

{

"url": "example.com",

"injection_detected": false,

"content_markdown": "# Example Domain…"

}

// how it works

no signup · no API key

01 / call

Hit the endpoint

Your agent calls any tool below over plain HTTP, or via the MCP server at /mcp/.

02 / settle

402, then one round-trip

You get an x402 payment challenge. Your client settles it in USDC on Base or Solana, automatically.

03 / receive

200, with your data

The retried request carries proof of payment and returns the real result. No accounts, no invoices.

// tools & pricing

$0.015 – $0.12 per call

x402 PAID

injection_scanPOST /tools/injection-scan $0.02
injection_scan_deepPOST /tools/injection-scan/deep $0.10
injection_scan_batchPOST /tools/injection-scan/batch $0.12
fetch_cleanPOST /tools/fetch-clean $0.015
fetch_clean_batchPOST /tools/fetch-clean/batch $0.10
safe_fetchPOST /tools/safe-fetch $0.03

settled instantly · USDCBase & Solana

SCANFETCHCLEANBATCH