Email Guard MCP Server

Agent Tools — deterministic tools for AI agents (MCP + APIs)

A family of deterministic tools that AI agents and developers call constantly — each exposed both as plain HTTP APIs and as an MCP server. No LLM in the loop, no API keys for the free tier, no tracking. Same input → same output. Just reliable, boring, useful tools.

The products

All four: deterministic, no LLM, free serverless tier; paid plans via RapidAPI for higher volume. The last three form the AI-agent safety suite: Package Guard (supply chain) · Agent Firewall (input/output) · Payment Guard (money).

Quick start (MCP)

Add any or all to your client's mcpServers config (Claude Desktop, Cursor, Claude Code, …):

{
  "mcpServers": {
    "agent-tools":     { "command": "npx", "args": ["-y", "web-tools-mcp"] },
    "package-guard":   { "command": "npx", "args": ["-y", "package-guard-mcp"] },
    "agent-firewall":  { "command": "npx", "args": ["-y", "agent-firewall-mcp"] },
    "payment-guard":   { "command": "npx", "args": ["-y", "payment-guard-mcp"] }
  }
}

---

1) Agent Web Tools — 10 web utilities · web-tools-mcp

Base: https://agent-tools-api.vercel.app. Code: agent-tools-mcp/ + agent-tools-api/.

2) Package Guard — supply-chain guard for coding agents · package-guard-mcp

verify_package (the pre-install guard), check_vulns (OSV), package_info, audit_deps, typosquat_scan. Data: OSV.dev + npm/PyPI. Base: https://package-guard.vercel.app. Code: package-guard-mcp/ + package-guard/.

3) Agent Firewall — input/output safety · agent-firewall-mcp

scan_content (prompt-injection/jailbreak/obfuscation), scan_secrets (+ PII redaction), check_url, check_ip, check_password (HIBP k-anonymity). Data: HIBP, RDAP, Tor, Team Cymru. Base: https://agent-firewall-seven.vercel.app. Code: agent-firewall-mcp/ + agent-firewall/.

4) Payment Guard — pre-send risk check for agents that move money · payment-guard-mcp

screen_address (address/ENS → OFAC-sanctioned? scam? on-chain risk → verdict), screen_payment (x402/ payment URL risk), check_sanctioned (fast OFAC), resolve_name (ENS → address, screened), screen_token (honeypot/rug/tax risk before a buy/approve, via on-chain simulation). Data: OFAC SDN + ethereum-lists + ScamSniffer + honeypot.is + public RPC + ENS. Chains: ETH/Base/Polygon/Arbitrum/Optimism. Base: https://payment-guard.vercel.app. Code: payment-guard-mcp/ + payment-guard/. x402-ready: an env-gated USDC micropayment gate is built into every endpoint (dormant until a wallet is set).

---

🛡️ Built right

http/https only · DNS-resolved SSRF guard · request timeouts · response size caps · content-type checks. Deterministic — same input, same output. No LLM, no paid data sources. Each API is a serverless function on a free tier; the MCP servers are thin stdio wrappers that call the same endpoints.

License

MIT — see LICENSE. Contributions and tool suggestions welcome.