Revolut Business MCP
Read-only Model Context Protocol (MCP) server exposing Revolut Business transaction data to MCP clients over stdio.
README
Revolut Business MCP
Read-only Model Context Protocol (MCP) server exposing Revolut Business transaction data to MCP clients over stdio.
Tools
revolut_list_transactions
List transactions. Single-page passthrough of the Revolut API response; no lossy field mapping.
| Parameter | Type | Notes |
|---|---|---|
from |
string (ISO calendar date or RFC3339 datetime) | optional |
to |
string (ISO calendar date or RFC3339 datetime) | optional |
account |
string (UUID) | optional, filter by account |
count |
integer 1–1000 | optional |
type |
enum | optional; one of atm, card_payment, card_refund, card_chargeback, card_credit, charge, charge_refund, exchange, transfer, loan, fee, refund, topup, topup_return, tax, tax_refund |
revolut_get_transaction
Get a single transaction.
| Parameter | Type | Notes |
|---|---|---|
id |
string | required |
id_type |
literal "request_id" |
optional; when set, id is treated as a request ID and only resolves transactions created within the last 2 weeks |
Pagination caveat
revolut_list_transactions returns one page only, matching the Revolut API's own behavior. The server does not auto-paginate. Use from/to to page through results yourself.
Prerequisites
- Node.js 22+
- A Revolut Business account with API access
- A client certificate registered with Revolut for API authentication, and the corresponding private key
Follow Revolut's official setup guides before configuring this server:
Install / build / test
npm install
npm run build # tsc -> dist/
npm run typecheck # tsc --noEmit
npm test # tsx --test test/**/*.test.ts
Environment configuration
Copy .env.example to .env and fill in values. Do not commit .env or any private key material.
.env is auto-loaded (via Node's --env-file-if-exists=.env) only by npm start and npm run dev. Running node dist/src/index.js directly does not load .env; supply the required variables externally (e.g. via your process manager or shell), or invoke Node yourself with --env-file-if-exists=.env.
| Variable | Description |
|---|---|
REVOLUT_ENV |
sandbox (default) or production. Selects the API base URL. |
REVOLUT_CLIENT_ID |
OAuth client ID from your Revolut Business API app. |
REVOLUT_JWT_ISS |
JWT iss claim value used when signing the client assertion. |
REVOLUT_PRIVATE_KEY |
PEM private key matching the certificate registered with Revolut. Supports literal \n line breaks (quoted single-line form), which the server normalises. |
REVOLUT_REFRESH_TOKEN |
Initial OAuth refresh token obtained via Revolut's authorization flow. |
REVOLUT_HTTP_TIMEOUT_MS |
Optional. HTTP request timeout in milliseconds, integer between 100 and 120000. Defaults to 15000. |
Sandbox vs production
REVOLUT_ENV=sandbox targets https://sandbox-b2b.revolut.com/api/1.0; REVOLUT_ENV=production targets https://b2b.revolut.com/api/1.0. Any other value fails config loading at startup.
Refresh-token behavior
The server exchanges the configured refresh token for a short-lived access token on first use, caches it in memory, and renews it automatically before expiry (60s skew) or on a 401 response. If Revolut rotates the refresh token in its response, the rotated token is adopted only in memory for the current process and is never written to .env or any file. On restart, the server always uses the REVOLUT_REFRESH_TOKEN value from the environment, regardless of any rotation that happened in a previous run. If the old token was invalidated by rotation, the process fails to authenticate until REVOLUT_REFRESH_TOKEN is updated.
Operators who need rotation to survive restarts must build their own secure external persistence (e.g., a secrets manager write-back) and update the environment before the next start. This server does not provide that mechanism.
Running locally (stdio)
npm run build
npm start
npm start runs node --env-file-if-exists=.env dist/src/index.js, which auto-loads .env if present. For development without a build step, npm run dev runs the same via tsx against src/index.ts with the same .env auto-loading.
The server communicates over stdio; it has no HTTP listener. Run it via an MCP client that spawns it as a subprocess. If your MCP client invokes node dist/src/index.js directly (bypassing the npm scripts), it will not auto-load .env; the client must supply the environment variables itself (see the config example below) or you must add --env-file-if-exists=.env to the invocation.
MCP client configuration example
Generic stdio MCP client config (adapt keys to your client's format):
{
"mcpServers": {
"revolut-business": {
"command": "node",
"args": ["/absolute/path/to/revolut-business-mcp/dist/src/index.js"],
"env": {
"REVOLUT_ENV": "sandbox",
"REVOLUT_CLIENT_ID": "...",
"REVOLUT_JWT_ISS": "...",
"REVOLUT_PRIVATE_KEY": "-----BEGIN PRIVATE KEY-----\n...\n-----END PRIVATE KEY-----",
"REVOLUT_REFRESH_TOKEN": "..."
}
}
}
}
Security notes
- Never commit
.env, private keys, or refresh tokens. - Treat the refresh token as a long-lived credential; rotate it if leaked.
- The private key must correspond to the certificate registered with Revolut for this API client; keep it out of shell history and logs.
- This server is read-only: it exposes no write, transfer, or payment-initiating tools.
Non-goals (current)
- No auto-pagination across multiple pages of transactions.
- No caching, persistence, or database layer.
- No write/transfer/payment tools — listing and single-transaction retrieval only.
- No transport other than stdio.
- No automated refresh-token persistence back to environment files.
This README does not claim live verification against the production Revolut API; behavior is based on the implemented code and Revolut's published API documentation.
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。