Satoshkin AML/KYT MCP Server
MCP server for AML/KYT crypto wallet screening, enabling AI clients to check wallet risk scores, source-of-funds tags, activity dates, and volume totals via Satoshkin backend.
README
Satoshkin AML/KYT MCP Server
MCP server for AML/KYT crypto wallet screening. It is designed to connect AI clients such as Claude Desktop, Claude Code, ChatGPT MCP clients, and internal agent tools to Satoshkin wallet risk checks.
This MCP server provides AI assistants with crypto compliance tools for AML (Anti-Money Laundering) and KYT (Know Your Transaction) screening. Compatible with Claude Desktop, Claude Code, and any MCP-compatible client.
With an API key configured, checks return real BitOK-backed data: risk score 0-100, risk level, source-of-funds tags, activity dates, and volume totals. Without a key the server runs in clearly-labeled mock mode.
Status
v0.2.0 performs real AML/KYT screening through the Satoshkin backend
(POST /api-accounts/aml-check, BitOK data) when SATOSHKIN_API_KEY is set.
Without an API key the server runs in mock mode: deterministic, clearly-labeled placeholder responses that can never be mistaken for a real verdict.
This repository does not contain BitOK credentials, production AML/KYT business logic, user data, or paid-tier enforcement. Those belong in the private backend service.
Available Tool
check_wallet_aml
Checks a crypto wallet address for AML/KYT risk. Addresses are validated against per-chain formats client-side before any quota is spent.
Input:
{
"address": "1A1zP1eP5QGefi2DMPTfTL5SLmv7DivfNa",
"chain": "BTC"
}
Supported chain values:
BTCETHTRONUSDT-ERC20USDT-TRC20
Real output (with SATOSHKIN_API_KEY):
{
"address": "1A1zP1eP5QGefi2DMPTfTL5SLmv7DivfNa",
"chain": "BTC",
"checked_at": "2026-06-10T12:00:00.000Z",
"risk_score": 17,
"risk_level": "low",
"tags": ["exchange"],
"first_activity": "2021-01-01",
"last_activity": "2026-06-01",
"total_volume_in_usd": "1000.00",
"total_volume_out_usd": "900.00",
"source": "bitok",
"cache_hit": false,
"rate_limit": { "remaining": 999, "resets_at": "2026-06-10T13:00:00.000Z" },
"balance": { "charged_usd": "0.55", "after_usd": "9.45" }
}
Errors come back as structured isError payloads with machine-readable codes:
unauthorized, insufficient_balance, rate_limited, validation_error,
bitok_upstream_error, bitok_timeout, upstream_unreachable,
invalid_response, internal_error, and http_<status> (fallback when a
backend error body has no error field). Every error payload also carries
an http_status field.
Results are informational risk indicators, not legal or compliance advice.
Configuration
| Env var | Default | Purpose |
|---|---|---|
SATOSHKIN_API_KEY |
(unset = mock mode) | API key sk_live_* / sk_test_* issued by Satoshkin |
SATOSHKIN_API_BASE_URL |
https://satoshkin.com |
Backend base URL (staging override) |
SATOSHKIN_TIMEOUT_MS |
30000 |
Request timeout; first-time checks may take several seconds |
Claude Desktop / Claude Code
This package is not published to npm yet. Use the public GitHub repository through npx.
Add this to your MCP settings:
{
"mcpServers": {
"satoshkin-aml": {
"command": "npx",
"args": ["-y", "github:satoshkin-dev/satoshkin-aml-mcp"],
"env": {
"SATOSHKIN_API_KEY": "sk_live_your_key_here"
}
}
}
}
Omit the env block to run in mock mode.
For local development, use the built entrypoint:
{
"mcpServers": {
"satoshkin-aml": {
"command": "node",
"args": ["/absolute/path/to/satoshkin-aml-mcp/dist/index.js"]
}
}
}
Local Development
git clone https://github.com/satoshkin-dev/satoshkin-aml-mcp.git
cd satoshkin-aml-mcp
npm install
npm run build
npm test
Run the server locally:
npm run dev
Run the built server:
npm run build
node dist/index.js
Check Scope
Shipped in v0.2:
- Risk score
0-100+ risk level (low / medium / high) - Source-of-funds tags
- First and last activity dates
- Total volume in and out (USD)
- Per-key rate limiting and billing through Satoshkin backend authentication
Planned next: counterparty and cluster analysis, allow/review/block decisions.
Monetization Model
The MCP server is a free open-source connector.
Real checks require a Satoshkin API key. Billing is per successful check ($0.55, debited from a prepaid key balance), the first real check per key is free, and cached repeat checks within 60 seconds are not charged. Keys are currently issued by Satoshkin support; a self-service dashboard is in development. Mock mode is free and unlimited.
Related Satoshkin Products
- P2P bot: https://satoshkin.com
- AML/KYT Telegram bot: https://t.me/SatoshkinKYTbot
- AML/KYT web app: planned
Use Cases
- OTC desks: Screen counterparty wallets before settling P2P trades
- Compliance officers: Quick wallet risk checks during AI-assisted workflows
- Crypto fund managers: Vet wallet addresses for sanctions/mixer exposure via AI agent
- Web3 builders: Add AML screening to Claude/ChatGPT-powered dApps
- Crypto journalists: Investigate wallet history with AI assistance
Why MCP Instead of REST API?
Traditional AML APIs (Chainalysis, Elliptic, TRM Labs) require:
- Custom integrations per AI client
- Backend infrastructure to broker calls
- Manual prompt engineering for context
This MCP server lets any AI assistant call AML checks natively, with structured input/output that the LLM understands. Install once, use across all your AI tools.
Roadmap
- [x] v0.1: Scaffold with mock checks
- [x] v0.2: Real BitOK integration via Satoshkin backend (per-check billing, free first check, mock mode without key)
- [ ] v0.3: Self-service key dashboard + top-up
- [ ] v0.4: Additional chains (Solana, BNB, Polygon)
- [ ] v0.5: Richer tag taxonomy and sanctions-specific flags
- [ ] v0.6: Cluster analysis and entity attribution
License
MIT
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。