MarketNow v5.0: We pivoted from marketplace to security infrastructure for AI agents

#mcp #ai #agents #security

The pivot

We just repositioned MarketNow. It is no longer an MCP marketplace.

It is security infrastructure for AI agents.

The marketplace is still there (9,248 skills, all free). But it is now the distribution layer, not the core product.

The core product is Sentinel.

What Sentinel does

Sentinel is a 10-layer security audit pipeline for MCP servers and agent tools:

L1.5 — Metadata analysis (auth, CORS, input validation, OAuth scopes)
L1.6 — Static analysis (Semgrep rules, secret detection, OSV dependency scan)
L1.7 — Malware pattern detection (binary launchers, suspicious install scripts)
L1.8 — Malware family signatures (48 YARA-equivalent rules)
L1.9 — Prompt injection screening (32 rules across 10 categories)
L2.5 — gVisor sandbox isolation (network=none, read-only, cap-drop ALL)
L3 — Runtime MCP Interceptor (real-time JSON-RPC guardrail)
ATC — Agent Trust Card (Ed25519 signed, RFC 8785 JCS)
x402 — Streaming metered billing ($0.01 USDC per call on Base)
A2A — Remote agent execution

The Trust API

This is the part that makes it infrastructure, not just a scanner:

curl https://marketnow.site/api/trust-score?skillId=mn-gen-00003

Returns:

{
  "trust_score": 8,
  "max_score": 10,
  "risk_level": "red",
  "recommendation": "safe_to_install",
  "certificate_url": "https://marketnow.site/api/audit-skill?certificate=1&skillId=mn-gen-00003",
  "layers_passed": { "l15": true, "l16": true, "l25": false }
}

Agents can consume this API before installing any tool. No need to re-run the audit locally.

1 million agents do not need to re-audit the same MCP. MarketNow does it once.

The Interceptor

A real-time JSON-RPC guardrail that sits between your agent and the MCP server:

curl -X POST https://marketnow.site/api/interceptor \
  -H "Content-Type: application/json" \
  -d x27{"jsonrpc":"2.0","method":"tools/call","params":{"name":"execute","arguments":{"command":"rm -rf /"}}}x27

Returns:

{
  "allowed": false,
  "decision": "block",
  "violations": [{
    "rule_id": "BLOCK_DANGEROUS_CMDS",
    "severity": "critical"
  }]
}

5 policy rules: block .env reads, block rm -rf, block process spawns, block system writes, warn on non-allowlisted network.

Security audit completed

We ran a full security audit on our own API:

ATC issuance: now requires Ed25519 proof-of-ownership + rate limited
ATC revocation: now requires CA secret (was open to anyone)
Stream payments: validates Ethereum wallet format + limits calls to 1000
Interceptor: 10KB payload limit (DoS protection)
Input sanitization: path traversal, XSS, command injection blocked

Stats (all verified real)

9,248 MCP skills catalogued
5,662 audited by Sentinel (score > 0)
257 gVisor sandbox runs completed
57 Ed25519-signed Agent Trust Cards issued
CA key: Ed25519 (RFC 8032, OID 1.3.101.112)
2 npm packages: marketnow-mcp v1.8.0 + marketnow-install-stack v1.1.0

Pricing

Tier	Price	What you get
Free	$0	Basic scan, trust score, public report
Developer	$49-99	Deep audit, signed report
Professional	$199-499	Runtime testing, Trust Card, re-audit
Continuous	$99-499/mo	Monitoring, CVE tracking, auto re-audit
Enterprise	$5k-50k+/yr	Private audits, API, SLA