MCP Open Finance

The Problem

AI agents can write code, analyze data, and chat. But they can't operate a business — collect payments, issue invoices, ship products, or notify customers. Especially not in Latin America, where every service has its own API, auth pattern, and quirks (mTLS for BACEN Pix v2, JWT-RSA for STP-SPEI, AFIP web services, contract-gated developer portals, and on).

Meanwhile, four agentic payment protocols are shipping in parallel — checkout protocols (Stripe ACP, Google UCP), authorization layers (AP2), and micropayment rails (x402) — and none of them compose cleanly with each other or with the regional rails businesses actually run on.

MCP Dev LATAM bridges both gaps. Traditional LATAM services + the new agentic payment protocols, all accessible through a single MCP interface.

The Solution

Each MCP server in this repo wraps a real provider — payments, fiscal, logistics, messaging, banking, ERP, identity, fraud, crypto, and the agentic protocols — so your agent can operate a complete business workflow.

🛒 Customer places order
  → 💳 Agent charges via Pix (Asaas / Mercado Pago / Zoop)
  → 📄 Agent issues NFe (Nuvem Fiscal / Focus NFe / NFe.io)
  → 📦 Agent generates shipping label (Melhor Envio / Skydropx)
  → 📱 Agent sends tracking via WhatsApp (Z-API / WhatsApp Cloud)
  → 📊 Agent records in ERP (Omie / Bling / Tiny)
  → 🏦 Agent reconciles balance (Stark Bank / Itaú / BTG)

Six systems. Zero human intervention. One agent.

Agentic Payment Protocols

Four servers that bridge the emerging agentic payment stack:

The Autonomy Spectrum

Each protocol sits at a different level of agent autonomy:

 Human-in-loop ◄──────────────────────────────► Fully autonomous

  ┌─────────┐   ┌─────────┐   ┌─────────┐   ┌─────────┐
  │   ACP   │   │   UCP   │   │   AP2   │   │   x402  │
  │ Stripe  │   │ Google  │   │ Google  │   │Coinbase │
  └─────────┘   └─────────┘   └─────────┘   └─────────┘
  User confirms   Commerce     User sets      No user.
  every purchase  lifecycle    rules, agent   Machine-to-
  in-chat         managed      acts within    machine at
                  by agent     budget/scope   HTTP layer

The Convergence Stack

┌─────────────────────────────────────────────┐
│  Application Layer        ACP / UCP         │  Chat UX, product discovery
├─────────────────────────────────────────────┤
│  Authorization Layer      AP2 / Mandates    │  Spend limits, audit trails
├─────────────────────────────────────────────┤
│  Tool Layer               MCP  ◄── WE ARE  │  Standardized agent tools
├─────────────────────────────────────────────┤
│  Settlement Layer         x402 / Pix / Card │  On-chain or traditional rails
└─────────────────────────────────────────────┘

MCP Dev LATAM sits at the Tool Layer — the middleware that connects every application, authorization, and settlement protocol through one interface.

Why this matters

Agent needs to buy something
  ├── Full commerce?       → Google UCP (search → cart → checkout → delivery)
  ├── Retail checkout?     → Stripe ACP (create_checkout → complete_checkout)
  ├── API micropayment?    → x402 (pay_request → USDC $0.001 → data returned)
  ├── Agent-to-agent?      → AP2 (authorize_payment → execute_payment)
  └── LATAM merchant?      → Asaas / Mercado Pago / Conekta / Wompi / etc.

All via MCP. Same interface. One agent.

Quick Start

With Claude Desktop

Add to ~/.config/claude/claude_desktop_config.json:

{
  "mcpServers": {
    "stripe-acp": {
      "command": "npx",
      "args": ["-y", "@codespar/mcp-stripe-acp"],
      "env": {
        "STRIPE_API_KEY": "sk_test_..."
      }
    },
    "asaas": {
      "command": "npx",
      "args": ["-y", "@codespar/mcp-asaas"],
      "env": {
        "ASAAS_API_KEY": "your-api-key",
        "ASAAS_SANDBOX": "true"
      }
    }
  }
}

For alpha packages (contract-gated providers like Itaú, Bradesco, Santander, Caixa, BB, BTG, C6, Sicoob, Bradesco, Matera), add @alpha to the install:

npx -y @codespar/mcp-itau@alpha

With any MCP client

npx @codespar/mcp-stripe-acp        # Agentic Commerce Protocol
npx @codespar/mcp-ucp                # Google Universal Commerce Protocol
npx @codespar/mcp-x402               # HTTP micropayments (Coinbase)
npx @codespar/mcp-ap2@alpha          # Agent-to-Agent payment authorization

npx @codespar/mcp-asaas              # BR billing + Pix
npx @codespar/mcp-mercado-pago       # LATAM payments
npx @codespar/mcp-nuvem-fiscal       # NFe / NFSe
npx @codespar/mcp-melhor-envio       # Multi-carrier shipping
npx @codespar/mcp-z-api              # WhatsApp messaging
npx @codespar/mcp-omie               # ERP
npx @codespar/mcp-stark-bank         # Banking
npx @codespar/mcp-brasil-api         # CEP, CNPJ (no key needed!)

Try it now (no API key)

BrasilAPI and BCRA are public — no key required. Try it in your terminal:

npx @codespar/mcp-brasil-api

Then ask your agent: "What is the address for CEP 01001-000?" or "Look up CNPJ 00.000.000/0001-91".

The Complete Loop

This is what makes the LATAM catalog different — not individual connectors, but a complete business workflow across verticals:

To orchestrate all six steps with governance, approval workflows, and audit trails — use CodeSpar.

All Servers

Browse the full catalog at codespar.dev/servers. Each server has its own README, env-var requirements, and tool reference under packages/<category>/<slug>/.

💳 Payments (40 servers)

🏦 Banking (12 servers)

📄 Fiscal (Brasil) (4 servers)

📱 Communication (8 servers)

📦 E-commerce / Logistics (7 servers)

📊 ERP / Accounting (5 servers)

🪪 Identity & KYC (5 servers)

🛡️ Fraud & Risk (4 servers)

🪙 Crypto / Stablecoins (9 servers)

🇦🇷 Argentina (5 servers)

🇨🇴 Colombia (5 servers)

🇲🇽 Mexico (6 servers)

Why MCP?

Model Context Protocol is the open standard for connecting AI agents to external tools. Instead of each agent building its own integrations, MCP provides a typed, discoverable interface that works with Claude, ChatGPT, Cursor, VS Code, and more.

AI Agent (Claude, ChatGPT, Cursor)
    ↕
MCP Server (this repo)
    ↕
LATAM API / Agentic Protocol (Stripe ACP, x402, Asaas, Mercado Pago, NFe.io, etc.)

Each MCP server in this repo:

• Exposes typed tools with input/output schemas
• Handles authentication (OAuth, mTLS, API keys, Basic Auth, JWT-RSA, signed requests)
• Supports dual transport — stdio (default) and Streamable HTTP (--http flag)
• Compatible with Claude Managed Agents via MCP Connector
• Supports sandbox mode for safe testing
• Returns structured JSON responses

Running in HTTP mode

Any server can run as an HTTP server for remote/cloud use:

# stdio (default — local, Claude Desktop, Cursor, Claude Code)
npx @codespar/mcp-asaas

# HTTP (remote — Managed Agents, cloud deployments)
npx @codespar/mcp-asaas --http
# or
MCP_HTTP=true npx @codespar/mcp-asaas

HTTP mode exposes /mcp (Streamable HTTP) and /health (status check).

Alpha vs stable

About a third of the catalog ships under the alpha npm dist-tag. These packages have correct tool schemas and auth flow, but the endpoint paths are best-guesses — typically because the provider's developer portal is contract-gated (BR public banks, AR central authority APIs, MX bank-transfer providers). They're flagged with TODO(verify) in source.

npm install @codespar/mcp-itau@alpha

Stable packages (the majority) install with the default tag:

npm install @codespar/mcp-stripe

Status is shown on every package page at codespar.dev/servers and in each package README.

About CodeSpar

CodeSpar builds commerce infrastructure for AI agents in Latin America — an MCP catalog (this repo), a runtime + SDK (@codespar/sdk on npm, codespar on PyPI), and a managed-tier governance product (AgentGate: programmable wallet, policy engine, CFO-grade audit, fiscal-compliance certifications).

Individual MCP servers are useful. Orchestrating many with governance is powerful. That's what CodeSpar does.

Contributing

We welcome contributions. See CONTRIBUTING.md.

Want a server for a service not listed? Open a server-request issue — there's a structured form with name, npm package, docs link, auth flow.

License

MIT — use freely in commercial and open source projects.