MarketNow — Security Infrastructure for AI Agents

Repo ecosystem (one owner per concern, split 2026-09-26): this repo owns the live marketplace — site (aep-marketplace/), catalog data (_data/, skills/, public/api/), the scheduled data pipelines and the Vercel deploys of marketnow.site. Product code (mcp-server = npm marketnow-mcp, atc-sdk, Docker/Cline/Cursor integrations) lives in alicelabs-llc/MARKETNOW; the ATC/1.0 protocol lives in alicelabs-llc/universal-trust-adapter.

MarketNow doesn’t sell AI tools. It determines whether AI agents should be allowed to trust and execute them.

What is MarketNow?

MarketNow is security infrastructure for AI agents. Not a marketplace.

The marketplace (71,198 MCP skills, all free) is distribution. The product is Sentinel — a 10-layer security audit pipeline that determines whether AI agents should be allowed to trust and execute tools.

Products

Sentinel — AI Agent Security Engine

10-layer audit pipeline:

Trust Card — Cryptographically verifiable identity

• Ed25519 signatures (RFC 8032)
• RFC 8785 JCS canonical JSON
• Public CA key: GET https://marketnow.site/api/atc?action=ca-key
• Verify any card: GET https://marketnow.site/api/atc?action=verify&card_id=ATC-2026-XXXXX

Interceptor — Real-time JSON-RPC guardrail

5 policy rules:

• Block reads of .env, .aws/credentials, .ssh/id_rsa
• Block dangerous commands (rm -rf, DROP TABLE, mkfs)
• Block process spawns (exec, spawn, child_process)
• Block system writes (/etc/, /root/, C:\Windows)
• Warn on non-allowlisted network calls

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

Trust API — Machine-readable trust decisions

curl https://marketnow.site/api/trust-score?skillId=mn-gen-00003
# → {"trust_score": 8, "recommendation": "safe_to_install"}

Stats (all verified real)

What Sentinel caught

3 quarantine decisions published (public ledger) for:

• Malware patterns (binary launchers, suspicious install scripts)
• Malware family signatures (48 YARA-equivalent rules matched)
• Prompt injection (32 rules, jailbreak attempts detected)
• Hacking tools (Kali Linux, darknet, OSINT attack frameworks)
• Credential exfiltration patterns
• Failed sandbox execution (container crashes, unresponsive)

Full transparency report: GET https://marketnow.site/api/audit-report.json

Quick start

Install MCP server

npx -y marketnow-mcp

Add to Cursor — 1 click (no terminal)

Or manually — ~/.cursor/mcp.json, claude_desktop_config.json, or any MCP client:

{
  "mcpServers": {
    "MarketNow": {
      "url": "https://marketnow.site/api/mcp"
    }
  }
}

Remote streamable-http, no auth required. Also listed in the official MCP Registry.

Install a skill stack

npx -y marketnow-install-stack security-analyst
npx -y marketnow-install-stack dev-productivity

Check trust score

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

Test the interceptor

curl -X POST https://marketnow.site/api/interceptor \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","method":"tools/call","params":{"name":"read_file","arguments":{"path":"/.env"}}}'

MCP Server v1.16.0 — Agent Contract

The npm package [email protected] exposes 15 tools, all under the marketnow_* namespace so Claude Desktop, Cursor, Cline, LangChain, and LlamaIndex can disambiguate them at tool-choice time.

The four golden rules enforced (full audit in mcp-server/AUDIT.md — product code lives in the MARKETNOW repo since the 2026-09-26 split):

The 15 tools:

1. marketnow_search_skills — keyword/category/price-bounded search
2. marketnow_get_skill — full skill detail by ID/slug
3. marketnow_list_categories — marketplace taxonomy with counts
4. marketnow_get_manifest — marketplace metadata + security metrics
5. marketnow_get_install_command — npx install command for a skill
6. marketnow_verify_trust — verify an Agent Trust Card (ATC)
7. marketnow_verify_receipt — verify a signed delivery proof (rcpt_*)
8. marketnow_submit_skill — submit a GitHub repo (L1.5+L1.7 sync, L2 queued)
9. marketnow_mint_referral — mint ref_xxxxxxxx (5% commission)
10. marketnow_lookup_referral — referral stats
11. marketnow_recommend_skills — AI-ranked recommendations for a task
12. marketnow_get_owasp_compliance — OWASP MCP Cheat Sheet (12 controls) + SHA-256 tool fingerprints + capability manifest
13. marketnow_check_revocation — check an Agent Trust Card or CA key against the signed Revocation Registry (MNR-CRL-1.0) + live ledger
14. marketnow_fingerprint_tool — RFC 8785 JCS + SHA-256 fingerprint of MCP tool definitions (drift/rug-pull detection)
15. marketnow_verify_atc_spec — verify any Agent Trust Card against the ATC/1.0 spec (10 controls, 8 required)

Pricing

Pricing coherence note (v5.0.0): MarketNow is security infrastructure, not a marketplace. The marketplace (71,198 MCP skills, all free to install) is distribution. The paid product is Sentinel — the security audit pipeline. The MCP server itself is free to install and use; paid actions kick in when you need a signed Trust Card, runtime testing, or continuous monitoring.

Links

• Website: https://marketnow.site
• GitHub: https://github.com/eddyflores100-lang/marketnow (this repo — live marketplace) · https://github.com/alicelabs-llc/MARKETNOW (product/protocol)
• npm: https://www.npmjs.com/package/marketnow-mcp
• MCP Server: npx -y marketnow-mcp
• Trust API: https://marketnow.site/api/trust-score
• Interceptor: https://marketnow.site/api/interceptor
• ATC Spec: https://marketnow.site/api/atc?action=spec
• CA Public Key: https://marketnow.site/api/atc?action=ca-key

License

AL code in this repository is PROPRIETARY — property of AliceLabs LLC.

For licensing: [email protected] For support: [email protected] General: [email protected]

Built by AliceLabs LLC (Wyoming, USA) — founder Edison Flores.

Taxonomy — which count belongs to which system

Three different counts coexist in this ecosystem. They are not three ways of counting the same thing:

If a surface says “8-layer audit” anywhere, it is stale — the Sentinel pipeline is 12 stages grouped into 10 audit layers (L1–L9). ATC’s “8” is the count of required verification controls (10 total). UTA’s number is formats, not layers.

Test CA keypair (cross-language reproducibility)

The test CA keypair is intentionally published (including private key) so any stranger can regenerate and re-sign the conformance vectors in any language.

⚠️ TEST ONLY — this private key is intentionally public. It MUST NEVER be trusted in production. ca-test-2 exists so any stranger can regenerate and re-sign the conformance vectors in any language. Signatures under ca-test-2 prove conformance-suite behavior — nothing else. Production CAs (mn-ca-003) are separate keys, never published, and their lifecycle is auditable in the revocation registry and the 2026-09-08 incident postmortem.