Mario A2MCP Intelligence Suite
Multi-endpoint Web3 intelligence service for token risk scanning, pre-trade checks, and signal snapshots, using DexScreener data with x402 payment support.
README
Mario A2MCP Intelligence Suite
Multi-endpoint A2MCP-style Web3 intelligence service for OKX.AI listing experiments.
The current suite exposes three separately listable services on one shared runtime:
| Service | Endpoint | Positioning |
|---|---|---|
| Token Risk Guard | POST /api/token-risk-scan |
Professional token risk scan for liquidity, volatility, data availability, and review hints. |
| ApeGuard | POST /api/ape-pretrade-check |
Short pre-trade meme/new-token check with an ape_score and decision hint. |
| Web3 Signal Snapshot | POST /api/signal-snapshot |
Generic Web3 signal snapshot for tokens, wallets, projects, or risk objects. |
The first production direction is Token Risk Guard plus ApeGuard. They share the same data, scoring, and x402 payment layer, so more A2MCP endpoints can be added without rebuilding payment and deployment from scratch.
Current Data Sources
- DexScreener token pairs API for public DEX liquidity, price, volume, pair age, and transaction-count fields.
- Holder concentration and contract-permission fields are explicitly returned as unavailable in this MVP. The service does not infer or fabricate them.
Run Locally
npm start
Then call:
curl -s http://localhost:8787/health
curl -s http://localhost:8787/metadata
curl -s http://localhost:8787/mcp
curl -s http://localhost:8787/openapi.json
Token Risk Guard:
curl -s -X POST http://localhost:8787/api/token-risk-scan \
-H 'content-type: application/json' \
-d '{
"chain": "solana",
"token_address": "So11111111111111111111111111111111111111112",
"language": "zh-CN"
}'
ApeGuard:
curl -s -X POST http://localhost:8787/api/ape-pretrade-check \
-H 'content-type: application/json' \
-d '{
"chain": "solana",
"token_address": "So11111111111111111111111111111111111111112",
"mode": "quick",
"language": "zh-CN"
}'
Signal Snapshot:
curl -s -X POST http://localhost:8787/api/signal-snapshot \
-H 'content-type: application/json' \
-d '{
"chain": "solana",
"address": "So11111111111111111111111111111111111111112",
"mode": "token",
"question": "给我一个代币风险和市场信号快照",
"lookbackHours": 24,
"language": "zh-CN"
}'
MCP-style tool call:
curl -s -X POST http://localhost:8787/mcp \
-H 'content-type: application/json' \
-d '{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": {
"name": "token_risk_scan",
"arguments": {
"chain": "solana",
"token_address": "So11111111111111111111111111111111111111112",
"language": "zh-CN"
}
}
}'
One-command Tests
The smoke tests start and stop a local server automatically.
npm test
npm run test:x402
Run Mock x402 Mode
This mode does not settle real payments. It verifies that unpaid requests return HTTP 402 plus a PAYMENT-REQUIRED header.
PORT=8788 \
PAYMENT_MODE=mock-x402 \
PUBLIC_BASE_URL=https://your-public-domain.example \
npm start
Then test any endpoint:
onchainos agent x402-check \
--endpoint http://localhost:8788/api/token-risk-scan \
--body '{"chain":"solana","token_address":"So11111111111111111111111111111111111111112"}'
onchainos agent x402-check \
--endpoint http://localhost:8788/mcp \
--body '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
Production x402 Mode
Production payment mode requires OKX facilitator credentials and a real receiving address. Do not enable this until the receiving wallet and OKX payment setup are confirmed.
PORT=8788 \
PAYMENT_MODE=okx-x402 \
OKX_API_KEY=... \
OKX_SECRET_KEY=... \
OKX_PASSPHRASE=... \
X402_PAY_TO=0xYourReceivingAddress \
X402_PRICE='$0.02' \
X402_NETWORK=eip155:196 \
X402_ASSET=0x779ded0c9e1022225f8e0630b35a9b54be713736 \
X402_AMOUNT_MINIMAL=20000 \
X402_TOKEN_NAME='USDt0' \
X402_TOKEN_SYMBOL=USDT0 \
X402_TOKEN_DECIMALS=6 \
npm start
X402_PRICE is kept for listing/metadata display. The actual payment challenge uses the explicit asset amount fields so validators can read token decimals without guessing.
Deployment Shape
This folder includes deployment starters:
.env.example: local and production environment variables.Dockerfile: container build for Node 22.14.deploy/systemd/a2mcp-signal-snapshot.service: Linux service template.deploy/nginx/a2mcp-signal-snapshot.conf: reverse proxy template before adding HTTPS.
For OKX.AI A2MCP listing, the endpoint should be a stable public HTTPS URL. Localhost and temporary tunnels are only useful for development checks.
Set PUBLIC_BASE_URL to the HTTPS origin so /metadata can include complete endpointUrl fields for each listable service.
Candidate OKX.AI A2MCP Listings
Token Risk Guard:
面向交易 Agent、研究 Agent 和钱包风控场景,按次返回结构化 Token 风险扫描结果,包括 risk_score、risk_level、flags、liquidity、holders、contract、suggested_action、data_quality 和 source URLs。第一版接入公开 DEX 数据源;未覆盖的数据会明确标记 unavailable,不编造。
Endpoint:
https://<your-domain>/api/token-risk-scan
MCP endpoint:
https://<your-domain>/mcp
ApeGuard:
面向 meme / 新币交易前体检场景,输入 chain 和 token_address,返回 ape_score、risk_level、one_line、red_flags、market_status 和 decision_hint。输出只做风险提示,不构成买卖建议,也不执行交易。
Endpoint:
https://<your-domain>/api/ape-pretrade-check
Suggested fee:
0.02 USDT / call
Production Checklist
- Deploy to a stable public HTTPS endpoint.
- Add more verifiable read-only data sources: explorer/security API for contract checks, holder distribution, and optional GoPlus/Honeypot providers.
- Keep missing-data fields explicit with
data_quality,available,reason, and source status. - Switch from
mock-x402tookx-x402with real OKX facilitator credentials and receiving address. - Add request logging without storing user secrets.
- Add rate limiting and timeout controls.
- Test with a real agent client before registering on OKX.AI.
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。