cabal-hunter-mcp
On-chain Solana cabal & rug detection as an MCP server. Scans any Solana token mint before your agent buys and returns an Exit-Liquidity Risk verdict.
README
cabal-hunter-mcp
On-chain Solana cabal & rug detection as an MCP server. One tool — check_cabal_risk — scans any Solana token mint before your agent buys and returns an Exit-Liquidity Risk verdict (SAFE | REVIEW | AVOID), a 0–100 cabal score, funding-cluster detection, same-block Jito-bundle detection, coordinated-dump detection, serial-rug deployer history ("launched 14, 13 dead"), and a Solana-native honeypot check (freeze authority + Token-2022 traps). Every flag links to its on-chain evidence transaction.
Contract-clean is not cabal-clean. RugCheck-style scanners tell you the mint/freeze/LP are fine — they don't tell you that 15 wallets funded from one source are holding 30% of supply, waiting to dump on you. That's what this catches.
Quick start
npx cabal-hunter-mcp
No install, no signup, no API key — 250 free scans/month per IP. That's the whole setup; the command below is what you drop into any MCP client.
Claude Desktop / Claude Code / Cursor / VS Code / ElizaOS
{
"mcpServers": {
"cabal-hunter": {
"command": "npx",
"args": ["-y", "cabal-hunter-mcp"]
}
}
}
Prefer a remote HTTP server (no local process)? Point straight at the hosted endpoint instead:
{ "mcpServers": { "cabal-hunter": { "url": "https://api.cabal-hunter.com/mcp" } } }
The tool
check_cabal_risk({ mint }) — pass a Solana token mint (contract) address. Returns the full forensic JSON:
{
"recommendation": "AVOID", // SAFE | REVIEW | AVOID ← the headline
"risk": "HIGH", // exit-liquidity risk
"cabal_score": 100, // 0-100
"honeypot_risk": "LOW", // freeze authority + Token-2022 traps
"mint_authority_revoked": true,
"freeze_authority_revoked": true,
"deployer": { "verdict": "SERIAL_RUGGER", "tokens_launched": 14, "dead": 13 },
"coordinated_clusters": [
{ "wallets": 5, "combined_pct": 23.1, "evidence_tx": "https://solscan.io/tx/…" }
],
"time_sync": true, // same-block (Jito-bundled) buys
"coordinated_exit": false, // ≥2 holders dumped together
"top_reasons": ["..."],
"wallets_checked": 15,
"scan_complete": true
}
scan_complete / wallets_checked are included on purpose so your agent can apply its own risk tolerance instead of inheriting ours — the score is a starting point you can verify (every cluster carries an evidence_tx), not a verdict you take on faith.
Gate a buy in your agent
"Before buying any token, call
check_cabal_riskwith the mint. IfrecommendationisAVOIDorcabal_score >= 65orhoneypot_riskisHIGH, skip the trade and say why."
Pricing
- 250 scans/month per IP — free, no key.
- After that: $9/month for Unlimited (fair use), or pay-as-you-go at $0.001 USDC per scan (priced at cost). No signup, no card.
- Prepaid key: send USDC,
POST /api/buy-key, then setCABAL_HUNTER_API_KEY(sent as theX-API-Keyheader). Full details: api.cabal-hunter.com/pricing.
Configuration
| Env var | Default | Purpose |
|---|---|---|
CABAL_HUNTER_API_KEY |
(none) | Prepaid key for unlimited / metered use (X-API-Key). |
CABAL_HUNTER_API |
https://api.cabal-hunter.com |
Override the API base URL. |
Other ways to use Cabal-Hunter
- REST:
curl "https://api.cabal-hunter.com/api/scan-cabal?mintAddress=<MINT>"— OpenAPI spec - ElizaOS plugin:
elizaos-plugin-cabal-hunter(npm install elizaos-plugin-cabal-hunter) - MCP template / starter: solana-safe-sniper-mcp-template
- Human? Free interactive bubble map — wallets, Solscan receipts, live chart + trade links on one screen: api.cabal-hunter.com/map
What it detects (why "contract-clean" misses it)
A cabal is 15 fresh wallets — all funded from the same master wallet, all buying in the first seconds of launch — quietly accumulating 25–40% of supply before your bot sees the first candle. Contract clean. LP burned. Everything green. Then they dump, simultaneously, into your liquidity. Cabal-Hunter traces the funding graph on-chain and answers the only question that matters before you sign a swap: are you the exit liquidity?
MIT licensed. Powered by Cabal-Hunter. This package is a thin MCP wrapper over the hosted API — the detection runs server-side against live Solana on-chain data (Helius RPC).
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。