dev.metamesh-uga/metamesh-uga

Official

by SiouxTXT · JavaScript

MetaMesh-UGA

Serverless edge control plane to discover, route and monetize 13,000+ MCP tools.

Edge control plane to discover, route, and monetize thousands of MCP tools.

mcp toolsdiscoveryroutingedge computingmonetizationserverless

Topics

agentic-paymentsai-agentscloudflare-workersllm-toolsmcpmcp-gatewaymcp-servermodel-context-protocolserverlessstripetool-usex402

README.md17 KBView on GitHub

MetaMesh-UGA

The MCP Operating System — A Serverless Control Plane for AI Agents and MCP Infrastructure

Set & Forget | Zero Infrastructure | Serverless | Edge Native | MCP Operating System

Live Status

Gateway: https://api.metamesh-uga.dev — operational
Dashboard: https://dashboard.metamesh-uga.dev
MCP endpoint: https://api.metamesh-uga.dev/mcp (JSON-RPC over HTTP)
MCP Registry: dev.metamesh-uga/metamesh-uga v2.0.0 — published & latest
Tools indexed: 13,000+ MCP servers, synced every 6h from the official MCP registry
Stripe billing: LIVE — agents top up a USD balance and are charged per call
x402 payments: implemented; activation gated by wallet config (see /v1/x402/info)

MetaMesh-UGA is the MCP Operating System: a serverless, edge-native control plane that discovers, verifies, scores, routes, monitors and scales MCP servers behind a single endpoint.

It combines registry sync, semantic discovery, trust scoring, policy enforcement, security scanning, lifecycle management, smart routing, multi-level caching, reliability patterns, self-healing and real-time analytics to make MCP infrastructure production-ready for AI agents.

Connect in seconds: point any MCP client (Claude, Windsurf, Cursor, Antigravity) at https://api.metamesh-uga.dev/mcp, or register an agent for a programmatic API key. Either way you gain access to a curated, trusted, and benchmarked catalog of MCP servers.

🚀 Caratteristiche

MCP Operating System: Control Plane + Data Plane per gestire l'intero ecosistema MCP
Unified Registry: Single source of truth con federation, mirroring e snapshot
Smart Discovery: Keyword, semantic search, capability graph e intent-based discovery
Trust Engine: Reputation score basato su uptime, latenza, success rate e security
Smart Routing: Weighted, latency, geographic, cost e health-based routing
Reliability Layer: Circuit breaker, retry, bulkhead, hedging e fallback
Multi-level Cache: In-memory L1 + KV L2, TTL, pattern invalidation
Security Fabric: Dependency scan, CVE detection, policy engine OPA/Rego
Self-Healing: Auto-deprecated, rollback recovery, health-based remediation
Analytics: Usage trends, cost analysis, performance metrics, export OTel/Prometheus
Lifecycle Management: Stati automatici dalla discovery all'archiviazione
Serverless: Zero infrastruttura da gestire (Cloudflare Workers)
WASM Runtime: Esecuzione sicura dei tool MCP compilati in WebAssembly
AI Agent Economy: Supporto nativo per x402 Protocol (micropagamenti USDC)
Monetizzazione: Piani Free, Pro ($19/mese), Enterprise ($499/mese)
Set & Forget: Automazione completa con cron triggers

📋 Requisiti

Node.js 18+
Go 1.21+ (per CLI)
Account Cloudflare (free tier sufficiente)
Account Stripe (per pagamenti)

🛠️ Installazione

1. Clone e Setup

bash

git clone https://github.com/SiouxTXT/metamesh-uga.git
cd metamesh-uga
npm install

2. Configurazione Cloudflare

bash

# Installa Wrangler
npm install -g wrangler

# Login a Cloudflare
wrangler login

# Crea database D1
wrangler d1 create metamesh-catalog

# Crea KV namespace
wrangler kv:namespace create "CACHE"

# Crea R2 bucket
wrangler r2 bucket create metamesh-wasm

3. Configura Secrets

bash

# Stripe
wrangler secret put STRIPE_SECRET_KEY
wrangler secret put STRIPE_WEBHOOK_SECRET

# JWT
wrangler secret put JWT_SECRET

# Alerting (opzionale)
wrangler secret put TELEGRAM_BOT_TOKEN
wrangler secret put TELEGRAM_CHAT_ID

# Admin
wrangler secret put ADMIN_KEY

4. Database Migration

bash

wrangler d1 execute metamesh-catalog --file=./shared/migrations/001_init.sql

5. Deploy

bash

# Deploy Workers
npm run deploy:workers

# Deploy Dashboard
cd packages/dashboard
npm install
npm run deploy

6. Installa CLI

bash

cd cli
go build -o metamesh

# Installa globalmente
mv metamesh /usr/local/bin/

# O usa direttamente
./metamesh connect

🎯 Uso

CLI Tool

bash

# Connetti al gateway (salva la config locale)
metamesh connect

# Chiama un tool (raggiunge POST /v1/call)
metamesh call example.echo --message "hello"

# Lista tool disponibili
metamesh list --category demo

# Visualizza utilizzo
metamesh usage

API REST

bash

# List tools (paginated, public)
curl "https://api.metamesh-uga.dev/v1/tools?limit=20&sort=popularity"

# Full-text search
curl "https://api.metamesh-uga.dev/v1/search?q=github"

# Execute a tool
curl -X POST \
  -H "Content-Type: application/json" \
  -d '{"tool": "example.echo", "params": {"message": "hello"}}' \
  https://api.metamesh-uga.dev/v1/call

MCP Protocol

Add to Devin / Cursor / Windsurf / Claude Desktop:

json

{
  "mcpServers": {
    "metamesh-uga": {
      "url": "https://api.metamesh-uga.dev/mcp",
      "transport": "http"
    }
  }
}

Or test directly via JSON-RPC:

bash

curl -X POST https://api.metamesh-uga.dev/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'

🤖 AI Agent Economy

Two complementary, production-ready payment models:

Model A — Stripe prepaid credit (recommended)

Agents top up a USD balance once via Stripe Checkout, then per-call pricing is debited automatically from the balance. No per-call fees, no crypto required.

bash

# 1. Register an agent (returns agent_id + api_key, shown once)
curl -X POST https://api.metamesh-uga.dev/v1/agent/register \
  -H "Content-Type: application/json" -d '{"name":"My Agent"}'

# 2. Top up via Stripe (returns a Checkout URL)
curl -X POST https://api.metamesh-uga.dev/v1/agent/topup \
  -H "X-Agent-Id: agent_xxx" -H "X-Agent-Key: ak_xxx" \
  -H "Content-Type: application/json" -d '{"amount_usd": 10}'

# 3. Call tools — balance is debited per call
curl -X POST https://api.metamesh-uga.dev/v1/call \
  -H "X-Agent-Id: agent_xxx" -H "X-Agent-Key: ak_xxx" \
  -H "Content-Type: application/json" \
  -d '{"tool":"example.echo","params":{"message":"hi"}}'

# 4. Check balance / history
curl https://api.metamesh-uga.dev/v1/agent/wallet -H "X-Agent-Id: agent_xxx" -H "X-Agent-Key: ak_xxx"
curl https://api.metamesh-uga.dev/v1/billing/info

Operator activation:

bash

wrangler secret put STRIPE_SECRET_KEY --env production       # sk_live_... or sk_test_...
wrangler secret put STRIPE_WEBHOOK_SECRET --env production   # whsec_... (Stripe webhook signing secret)
# Point a Stripe webhook to: https://api.metamesh-uga.dev/v1/stripe/webhook (event: checkout.session.completed)
# Optionally set BILLING_ENABLED="true" in wrangler.toml to require payment for ALL callers.

When Stripe is unset, anonymous calls stay free and /v1/agent/topup returns a clear 503.

Current production state: Stripe (live keys) and the checkout.session.completed webhook are configured, and BILLING_ENABLED="true". Enforcement applies to the REST POST /v1/call endpoint — every caller is charged per request, so top up via /v1/agent/topup first. The public MCP endpoint (POST /mcp) stays free for discovery (tools/list) and native metamesh.* tools, so MCP clients keep working without an account.

Model B — x402 crypto (USDC on Base)

The x402 protocol: agents pay per call in USDC on-chain. Send an X-PAYMENT header; on HTTP 402 read the accepts requirements.

bash

curl https://api.metamesh-uga.dev/v1/x402/info

Payment flow

Call a paid resource (POST /v1/call). When enforcement is active and no payment is attached, the gateway responds HTTP 402 with a standard accepts requirements body (scheme exact, network base, asset USDC, payTo, maxAmountRequired).
The agent constructs a signed payment and retries with an X-PAYMENT header (base64 JSON).
The gateway verifies + settles via the configured facilitator, then returns the result plus an X-PAYMENT-RESPONSE header with the on-chain tx.

Activation (operator)

x402 is implemented but disabled until you set a receiving wallet:

bash

# In wrangler.toml [env.production.vars]
X402_ENABLED = "true"
X402_PAY_TO  = "0xYourBaseWallet"
# X402_FACILITATOR_URL and USDC_CONTRACT have working defaults

Per-tool pricing lives in the tool_pricing table (default $0.001/call, bulk $0.0005).

📊 Architettura

MetaMesh-UGA adotta un'architettura a due piani ispirata ai service mesh moderni.

code

┌─────────────────────────────────────────────────────────────────────────────┐
│                      META MESH-UGA — MCP OPERATING SYSTEM                 │
├─────────────────────────────────────────────────────────────────────────────┤
│                                                                             │
│  ┌─────────────────────────────────────────────────────────────────────┐   │
│  │                     CONTROL PLANE                                   │   │
│  │                                                                     │   │
│  │  ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌──────────┐ │   │
│  │  │ Registry │ │ Discovery│ │  Trust   │ │  Policy  │ │ Lifecycle│ │   │
│  │  │  Engine  │ │  Engine  │ │  Engine  │ │  Engine  │ │  Manager │ │   │
│  │  └──────────┘ └──────────┘ └──────────┘ └──────────┘ └──────────┘ │   │
│  │  ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌──────────┐             │   │
│  │  │ Security │ │Benchmark │ │Analytics │ │  Config  │             │   │
│  │  │  Scanner │ │  Engine  │ │  Engine  │ │  Engine  │             │   │
│  │  └──────────┘ └──────────┘ └──────────┘ └──────────┘             │   │
│  └─────────────────────────────────────────────────────────────────────┘   │
│                                    │                                        │
│                                    ▼                                        │
│  ┌─────────────────────────────────────────────────────────────────────┐   │
│  │                       DATA PLANE                                    │   │
│  │                                                                     │   │
│  │  ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌──────────┐ │   │
│  │  │  Routing │ │   Cache  │ │  Health  │ │  Proxy   │ │  Rate    │ │   │
│  │  │  Engine  │ │  Engine  │ │  Engine  │ │  Engine  │ │ Limiting │ │   │
│  │  └──────────┘ └──────────┘ └──────────┘ └──────────┘ └──────────┘ │   │
│  └─────────────────────────────────────────────────────────────────────┘   │
│                                    │                                        │
│                                    ▼                                        │
│  ┌─────────────────────────────────────────────────────────────────────┐   │
│  │                         MCP ECOSYSTEM                               │   │
│  │  (Server MCP A, B, C, D, E...)                                     │   │
│  └─────────────────────────────────────────────────────────────────────┘   │
└─────────────────────────────────────────────────────────────────────────────┘

Control Plane: decisioni, configurazione, orchestrazione, analytics.
Data Plane: esecuzione delle richieste, routing, caching, health in tempo reale.

Per maggiori dettagli: docs/ARCHITECTURE.md.

💰 Piani di Pagamento

Piano	Prezzo	Feature
Free	€0	1.000 chiamate/mese, 100 req/min
Pro	$19/mese	Chiamate illimitate, 1.000 req/min
Enterprise	$499/mese	Dedicated VPC, 10.000 req/min, SLA 99.9%

📁 Struttura Repository

code

metamesh-uga/
├── packages/
│   ├── gateway/            # Data Plane: API Gateway, Routing, Cache, Proxy, Reliability
│   ├── discovery/          # Control Plane: Discovery Engine
│   ├── registry/           # Control Plane: Registry Engine (federation, mirroring, snapshot)
│   ├── trust/              # Control Plane: Trust Engine
│   ├── security/           # Control Plane: Security Scanner
│   ├── policy/             # Control Plane: Policy Engine
│   ├── benchmark/          # Control Plane: Benchmark Engine
│   ├── lifecycle/          # Control Plane: Lifecycle Manager
│   ├── recommendation/     # Control Plane: Recommendation Engine
│   ├── cost/               # Control Plane/Data Plane: Cost Optimizer
│   ├── compatibility/      # Control Plane: Compatibility Engine
│   ├── intelligence/       # Control Plane: AI Intelligence Layer
│   ├── analytics/          # Control Plane: Analytics Engine
│   ├── health/             # Data Plane: Health Engine
│   ├── config/             # Control Plane: Config Engine
│   ├── cache/              # Data Plane: Advanced Cache Engine
│   ├── self-healing/       # Data Plane: Self-Healing Engine
│   ├── aggregator/         # Motore di aggregazione
│   ├── inserter/           # Motore di inserimento
│   ├── updater/            # Motore di aggiornamento
│   ├── eliminatore/        # Motore di eliminazione
│   ├── alerts/             # Alerting Worker
│   ├── agent-billing/      # Billing per AI Agent
│   ├── dashboard/            # Cloudflare Pages (React)
│   └── landing/            # Landing page
├── cli/                    # CLI Tool (Go)
├── shared/                 # Librerie condivise
│   ├── src/
│   │   ├── auth.js         # Autenticazione
│   │   ├── ratelimit.js    # Rate limiting
│   │   ├── wasm-runtime.js # WASM execution
│   │   └── x402.js         # x402 Protocol
│   └── migrations/         # Database migrations
├── docs/                   # Documentazione
└── wrangler.toml           # Configurazione Cloudflare

🔧 Configurazione Environment

bash

# .env
CLOUDFLARE_API_TOKEN=xxx
CLOUDFLARE_ACCOUNT_ID=xxx
D1_DATABASE_ID=xxx
JWT_SECRET=xxx
STRIPE_SECRET_KEY=sk_test_xxx
STRIPE_WEBHOOK_SECRET=whsec_xxx
TELEGRAM_BOT_TOKEN=xxx
TELEGRAM_CHAT_ID=xxx

🧪 Testing

bash

# Test locale
wrangler dev --config packages/gateway/wrangler.toml

# Test API
curl http://localhost:8787/health

📝 Licenza

MIT License - vedi LICENSE

👤 Autore

Davanzo Keoma - GitHub

Il "set & forget" è realtà. 🚀

Install

claude_desktop_config.json

{
  "mcpServers": {
    "metamesh-uga": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://api.metamesh-uga.dev/mcp"
      ]
    }
  }
}

Desktop config is stdio-only; this bridges via mcp-remote. Native remote: Settings > Connectors.