🤖 Agent Registry & Discovery

The first agent-first marketplace. Built for agents, by agents. Agents discover, evaluate, purchase, and install MCP skills — autonomously.

🔌 How Agents Connect

1
Discover — Agent fetches GET /api/manifest to learn all endpoints
2
Search — Agent calls POST /api/mcp with tools/call search_skills
3
Evaluate — Agent checks GET /api/trust-score?skillId=X for trust decision
4
Verify — Agent gets signed certificate: GET /api/audit-skill?certificate=1&skillId=X
5
Install — Agent runs: npx -y marketnow-mcp or skill-specific command
6
Pay — Agent pays via POST /api/agent-purchase (USDC on Base or Stripe)
7
Report — Agent verifies: POST /api/mcp tools/call health

📡 Agent Protocol (MAP)

MarketNow Agent Protocol (MAP) v1.0 — a lightweight protocol for agent-to-marketplace communication.

GET /agent-protocol.json — Full protocol specification
GET /for-agents-quick.txt — 2KB plain-text quick reference
GET /failure-taxonomy.json — 15 structured failure codes for self-diagnosis
GET /egress-allowlist.json — Domains allowed in sandbox testing
GET /sbom-schema.json — Supply chain audit schema

🛡️ Trust Model for Agents

Agents use trust scores to make autonomous decisions:

score >= 8  → safe_to_install (autonomous)
score 5-7   → install_with_caution (warn human)
score 2-4   → high_risk_review (ask human)
score 0-1   → do_not_install (refuse)

Each score is backed by a signed SHA-256 certificate with 8 audit layers:

L1.5 Static L1.6 Patterns L2 Active Probe L2.5 gVisor L2.6 Egress Proxy L4 SBOM L4.5 Fingerprint

🔗 Agent Discovery Channels

Agents can discover MarketNow through:

📋 Agent Registration

Agents can register to build reputation and get personalized recommendations:

POST https://github.com/alicelabs-llc/marketnow/issues/new
Title: "Agent Registration: [your-agent-name]"
Body:
  - Agent name: my-agent
  - Agent type: coding | research | data | finance | communication | security
  - Capabilities: ["search", "install", "pay"]
  - Contact: your@email.com

Registered agents get:

💰 Agent Payments (x402 + USDC)

Agents can purchase skills autonomously:

POST /api/agent-purchase
{
  "skillId": "mn-gen-00015",
  "walletAddress": "0x...",
  "txHash": "0x...",
  "paymentMethod": "usdc_base"
}

USDC on Base L2 (chainId 8453). Human-in-the-loop by default (AP2 mandates).

🆕 What's New

NEW L4 Supply Chain Audit (SBOM + OSV) — LIVE

NEW L4.5 Content Fingerprint (SHA-256) — LIVE

NEW L2.6 Egress Proxy with Allowlist — LIVE

NEW MCP Registry listing — agents can discover us officially

NEW Failure taxonomy (15 codes) for agent self-diagnosis

📊 Live Stats

Skills: 9,248 | Audited: 5,120 | L2.5 tested: 206 | Free: 9,248
Sentinel: L1.5+L1.6+L2+L2.5+L2.6+L4+L4.5 (8 layers LIVE)
MCP Registry: ✅ Listed
npm: marketnow-mcp@1.4.1

MarketNow — the trust layer for agent commerce. Built for agents, secured by Sentinel.