streamable-httpupdated 2mo ago
The MCP Operating System β A Serverless Control Plane for AI Agents and MCP Infrastructure
What can you do with MetaMesh UGA?
MetaMesh-UGA
The MCP Operating System β 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.devβ 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-ugav2.0.0 β published & latest - Tools indexed: 13,000+ MCP servers, synced every 6h from the official MCP registry
- Stripe billing: LIVE β 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.
π 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
π Requisiti
- Node.js 18+
- Go 1.21+ (per CLI)
- Account Cloudflare (free tier sufficiente)
- Account Stripe (per pagamenti)
π οΈ Installazione
1. Clone e Setup
git clone https://github.com/SiouxTXT/metamesh-uga.git
cd metamesh-uga
npm install
2. Configurazione Cloudflare
# 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
# 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
wrangler d1 execute metamesh-catalog --file=./shared/migrations/001_init.sql
5. Deploy
# Deploy Workers
npm run deploy:workers
# Deploy Dashboard
cd packages/dashboard
npm install
npm run deploy
6. Installa CLI
cd cli
go build -o metamesh
# Installa globalmente
mv metamesh /usr/local/bin/
# O usa direttamente
./metamesh connect
π― Uso
CLI Tool
# 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
# 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:
{
"mcpServers": {
"metamesh-uga": {
"url": "https://api.metamesh-uga.dev/mcp",
"transport": "http"
}
}
}
Or test directly via JSON-RPC:
curl -X POST https://api.metamesh-uga.dev/mcp \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'
π€ AI Agent Economy
Two complementary, production-ready payment models:
Model A β 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.
# 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 β 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:
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.completedwebhook are configured, andBILLING_ENABLED="true". Enforcement applies to the RESTPOST /v1/callendpoint β every caller is charged per request, so top up via/v1/agent/topupfirst. The public MCP endpoint (POST /mcp) stays free for discovery (tools/list) and nativemetamesh.*tools, so MCP clients keep working without an account.
Model B β 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.
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 standardacceptsrequirements body (schemeexact, networkbase, asset USDC,payTo,maxAmountRequired). - The agent constructs a signed payment and retries with an
X-PAYMENTheader (base64 JSON). - The gateway verifies + settles via the configured facilitator, then returns the result plus an
X-PAYMENT-RESPONSEheader with the on-chain tx.
Activation (operator)
x402 is implemented but disabled until you set a receiving wallet:
# 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).
π Architettura
MetaMesh-UGA adotta un'architettura a due piani ispirata ai service mesh moderni.
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β META MESH-UGA β MCP OPERATING SYSTEM β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ€
β β
β βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ β
β β CONTROL PLANE β β
β β β β
β β ββββββββββββ ββββββββββββ ββββββββββββ ββββββββββββ ββββββββββββ β β
β β β Registry β β Discoveryβ β Trust β β Policy β β Lifecycleβ β β
β β β Engine β β Engine β β Engine β β Engine β β Manager β β β
β β ββββββββββββ ββββββββββββ ββββββββββββ ββββββββββββ ββββββββββββ β β
β β ββββββββββββ ββββββββββββ ββββββββββββ ββββββββββββ β β
β β β Security β βBenchmark β βAnalytics β β Config β β β
β β β Scanner β β Engine β β Engine β β Engine β β β
β β ββββββββββββ ββββββββββββ ββββββββββββ ββββββββββββ β β
β βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ β
β β β
β βΌ β
β βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ β
β β DATA PLANE β β
β β β β
β β ββββββββββββ ββββββββββββ ββββββββββββ ββββββββββββ ββββββββββββ β β
β β β Routing β β Cache β β Health β β Proxy β β Rate β β β
β β β Engine β β Engine β β Engine β β Engine β β Limiting β β β
β β ββββββββββββ ββββββββββββ ββββββββββββ ββββββββββββ ββββββββββββ β β
β βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ β
β β β
β βΌ β
β βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ β
β β MCP ECOSYSTEM β β
β β (Server MCP A, B, C, D, E...) β β
β βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
- Control Plane: decisioni, configurazione, orchestrazione, analytics.
- Data Plane: esecuzione delle richieste, routing, caching, health in tempo reale.
Per maggiori dettagli: docs/ARCHITECTURE.md.
π° Piani di Pagamento
| Piano | Prezzo | Feature |
|---|---|---|
| Free | β¬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% |
π Struttura Repository
metamesh-uga/
βββ packages/
β βββ gateway/ # Data Plane: API Gateway, Routing, Cache, Proxy, Reliability
β βββ discovery/ # Control Plane: Discovery Engine
β βββ registry/ # Control Plane: Registry Engine (federation, mirroring, snapshot)
β βββ trust/ # Control Plane: Trust Engine
β βββ security/ # Control Plane: Security Scanner
β βββ policy/ # Control Plane: Policy Engine
β βββ benchmark/ # Control Plane: Benchmark Engine
β βββ lifecycle/ # Control Plane: Lifecycle Manager
β βββ recommendation/ # Control Plane: Recommendation Engine
β βββ cost/ # Control Plane/Data Plane: Cost Optimizer
β βββ compatibility/ # Control Plane: Compatibility Engine
β βββ intelligence/ # Control Plane: AI Intelligence Layer
β βββ analytics/ # Control Plane: Analytics Engine
β βββ health/ # Data Plane: Health Engine
β βββ config/ # Control Plane: Config Engine
β βββ cache/ # Data Plane: Advanced Cache Engine
β βββ self-healing/ # Data Plane: Self-Healing Engine
β βββ aggregator/ # Motore di aggregazione
β βββ inserter/ # Motore di inserimento
β βββ updater/ # Motore di aggiornamento
β βββ eliminatore/ # Motore di eliminazione
β βββ alerts/ # Alerting Worker
β βββ agent-billing/ # Billing per AI Agent
β βββ dashboard/ # Cloudflare Pages (React)
β βββ landing/ # Landing page
βββ cli/ # CLI Tool (Go)
βββ shared/ # Librerie condivise
β βββ src/
β β βββ auth.js # Autenticazione
β β βββ ratelimit.js # Rate limiting
β β βββ wasm-runtime.js # WASM execution
β β βββ x402.js # x402 Protocol
β βββ migrations/ # Database migrations
βββ docs/ # Documentazione
βββ wrangler.toml # Configurazione Cloudflare
π§ Configurazione Environment
# .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
π§ͺ Testing
# Test locale
wrangler dev --config packages/gateway/wrangler.toml
# Test API
curl http://localhost:8787/health
π Licenza
MIT License - vedi LICENSE
π€ Autore
Davanzo Keoma - GitHub
Il "set & forget" Γ¨ realtΓ . π
Install
Add MetaMesh UGA to your client. Pick the one you use.
claude mcp add --transport http metamesh-uga https://api.metamesh-uga.dev/mcpcodex mcp add metamesh-uga --url https://api.metamesh-uga.dev/mcp{
"mcpServers": {
"metamesh-uga": {
"url": "https://api.metamesh-uga.dev/mcp"
}
}
}Add to `~/.cursor/mcp.json`, or `.cursor/mcp.json` for a single project.
{
"servers": {
"metamesh-uga": {
"type": "http",
"url": "https://api.metamesh-uga.dev/mcp"
}
}
}Add to `.vscode/mcp.json` in your workspace.
{
"mcpServers": {
"metamesh-uga": {
"url": "https://api.metamesh-uga.dev/mcp"
}
}
}Add to `claude_desktop_config.json`, then restart Claude Desktop.
{
"mcpServers": {
"metamesh-uga": {
"serverUrl": "https://api.metamesh-uga.dev/mcp"
}
}
}Add to `~/.codeium/windsurf/mcp_config.json`.
Score
39 / 100
Incomplete
- Documentation25/25
- Maintenance22/25
- Trust6/20
- Capability0/15
- Install experience12/15
- Documents what it does and how to connect
- Has a resolvable package or endpoint
- Exposes at least one tool, prompt or resource
- README has substantive content
- Includes a code example
- Documents its configuration
- Mentions credentials or security posture
- Last commit 61 days ago
- Has a release history
- Repository is not archived
- No licence detected
- Namespace verified in the official MCP registry
- Claimed by its owner
- Published under an organisation
- 0 tool(s) documented
- Provides prompt templates
- Provides resources
- 6 documented install method(s)
- Published to a package registry
- Offers a hosted endpoint β no local install
Version history
| Versions | Published |
|---|---|
| 2.0.3Latest | Jul 9, 2026 |
| 2.0.2 | Jul 5, 2026 |
| 2.0.1 | Jun 29, 2026 |
| 2.0.0 | Jun 28, 2026 |
| 1.0.0 | Jun 24, 2026 |