Pulsefeed X402

Client-side, MIT-licensed tools for PulseFeed — the independent trust & safety layer for the x402 agent-payment economy.

Before an AI agent pays an x402 endpoint, it should know whether that endpoint is safe to pay — ~70% of x402 endpoints are dead, invalid, or scams, and only about half of what catalogs call "healthy" actually work. PulseFeed independently probes every x402 service, scans for scams (payTo hijack, bait-and-switch, honeypot), verifies the receiver on-chain, and publishes an open Trust Score.

This repo holds the open-source client packages. They talk to PulseFeed's public API (free /verify, live /status, paid /trust). The crawler, scoring engine, and dataset are the service itself at pulsefeed.dev.

Packages

Package	What it does	Install
pulsefeed-x402-ai-tools	Drop-in tools for the Vercel AI SDK and LangChain: verify an x402 endpoint before your agent pays + discover working services	npm i pulsefeed-x402-ai-tools
pulsefeed-x402-guard	Wrap your paying fetch — every payment is verified first; dead/scam endpoints are blocked automatically (zero deps)	npm i pulsefeed-x402-guard
pulsefeed-x402-mcp	MCP server: gives Claude Desktop / Cursor / Cline tools to find live x402 services and verify endpoints before paying	npx -y pulsefeed-x402-mcp

Quick example

import { verifyX402Endpoint } from "pulsefeed-x402-ai-tools";

const v = await verifyX402Endpoint("https://api.some-x402-service.com/data");
if (v.verdict === "avoid") throw new Error("don't pay this endpoint");

MCP config

{ "mcpServers": { "pulsefeed": { "command": "npx", "args": ["-y", "pulsefeed-x402-mcp"] } } }

MCP server safety

Beyond x402, PulseFeed applies the same independent-audit playbook to the MCP server ecosystem — because a bad MCP server can run arbitrary code on your machine the moment you install it, and most people connect one without ever checking.

• 🔎 MCP Observatory — independent safety audit of the MCP server ecosystem (install-script / code-execution risk, abandonment, provenance, license, liveness)
• 📊 State of MCP Security — daily report. Right now ~10% of ~800 audited MCP servers run an install script (arbitrary code at npm i), and dozens ship no repo to review. (Install scripts aren't always malicious — native builds use them — but each is an unreviewed code-execution vector.)
• ✅ Check any server free: GET https://pulsefeed.dev/mcp/verify?package=<npm-name> — verdict + flags before you connect it

Security Disclosures & Data API

• 🚨 Security Disclosures — a live feed of real threats caught in the x402 economy (receiver hijacks, bait-and-switch pricing, honeypots, unverified receivers), each with on-chain proof you can verify yourself. Machine feed: /incidents.json · RSS
• 📦 Data API — the wholesale trust backbone for agentic commerce: every x402 service’s moat profile + every audited MCP server + ecosystem aggregates + live incidents in one call ($1 over x402). Free preview + full schema: /data.json

Links

• 🌐 Live observatory: pulsefeed.dev/status — the live state of x402
• 🧩 MCP Observatory · State of MCP Security — is that MCP server safe to install?
• 📊 State of x402 report · Trust Score standard · Change feed
• 🚨 Security Disclosures (RSS) · 📦 Data API — cross-domain trust dataset in one call
• 📖 Methodology · OpenAPI · llms.txt

MIT © PulseFeed