Fin-MCP Payment Server
Enables AI agents to interact with multiple payment providers (Stripe, Paystack) through a unified API. Supports payment initialization, verification, refunds, customer management, and invoicing without requiring knowledge of specific provider implementations.
README
Fin-MCP
Fin-MCP is a self-hosted, modular MCP that allows AI agents or enterprise systems to interact with multiple payment providers (e.g., Stripe, Paystack) via a single unified API. It abstracts the provider logic so agents can call payment operations seamlessly.
Features
- MCP for Payments: AI agents can perform payment actions without knowing provider details.
- Dynamic MCP tools: Initialize payment, verify payment, refund, create customers, and invoices.
- Optional authentication: Protect MCP access with API keys.
- Metadata endpoint: Agents can discover available tools and supported providers.
- Self-hosted MCP: Full control over deployments and provider credentials.
- Extensible: Add new tools or providers with minimal changes.
- Protocol-driven tool execution: Agents can call any tool via
/mcp/call. - Provider abstraction: MCP selects or routes to the appropriate payment provider automatically (e.g., based on provider like stripe, based on currency (soon)).
Table of Contents
- Installation
- Configuration
- Running the MCP
- API Endpoints
- Adding New Providers
- Development Notes
- License
Installation
git clone https://github.com/Veri5ied/fin-mcp.git
cd fin-mcp
Install dependencies
bun install
Configuration
Create a .env file in the root:
# Enable or disable API key authentication
ENABLE_AUTH=false
# API key for auth (only used if ENABLE_AUTH=true)
FIN_MCP_KEY=supersecret-fin-key
# Stripe secret key (required for Stripe provider)
STRIPE_SECRET=sk_test_1234567890
ENABLE_AUTH: Set totruefor secure access;falsefor testing or private networks.FIN_MCP_KEY: Your self-generated secret key.STRIPE_SECRET: Stripe API secret for payment operations.
Running the MCP
bun run dev
Open http://localhost:3000 in your browser.
API Endpoints
1. Get Metadata
GET /mcp/metadata
Agents discover available payment tools and providers
{
"tools": [
{
"name": "initialize_payment",
"description": "Start a payment",
"providers": ["stripe", "paystack"]
}
]
}
2. Call a Tool
POST /mcp/call
Agents execute payment operations through the MCP protocol.
Headers (if auth enabled):
Authorization: Bearer <FIN_MCP_KEY>
Content-Type: application/json
Body Example (Stripe Initialize Payment):
{
"tool": "initialize_payment",
"input": {
"provider": "stripe",
"amount": 5000,
"currency": "USD",
"email": "customer@example.com"
},
"idempotency_key": "unique-key-123"
}
Response:
{
"status": "ok",
"result": {
"id": "cs_test_1234",
"object": "checkout.session"
}
}
Adding New Providers
- Create a new adapter in
src/providers/. - Implement the required interfaces (
PaymentProvider, or extend specific tool interfaces). - Add the adapter to the provider registry (
src/providers/registry.ts). - Update
metadata.tsto include the new provider for relevant tools.
The MCP will automatically route calls based on the
providerfield in the request payload.
Development Notes
- Idempotency: Use
idempotency_keyin/mcp/callto prevent duplicate operations. - Audit Logging: MCP logs every call via
logAudit()for traceability. - Extensibility: You can define new tools without modifying the core callHandler — just add the tool interface and adapter implementation.
- Optional Authentication: Set
ENABLE_AUTH=falsefor testing or private self-hosted environments.
License
MIT License — free to use and modify for your self-hosted deployments.
💡 Optional Add-ons
- Integrate AI agents to call
/mcp/calldynamically. - Add support for webhooks from payment providers.
- Add fully typed output interfaces for each provider to eliminate
anyresponses.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。