CoinLobster MCP
Provides real-time crypto whale trade data and market analysis to AI agents, including unusual flow radar, liquidations, funding rates, and market snapshots across 15 exchanges and on-chain DEXs.
README
CoinLobster MCP · Live Crypto Whale Trades for AI Agents
No API key required. Real executed whale trades (not wallet transfers) merged live across 15 exchanges (Binance, Coinbase, Bybit, OKX, Kraken and more) plus on-chain DEX swaps on Ethereum, Base and Arbitrum, with a Smart Money Radar for unusual flow, liquidations, market snapshots and outcome-scored signals.
The only MCP server with live cross-exchange whale trades. Price data is everywhere, what whales just did is not.
Try asking
- "What are whales doing with BTC right now?"
- "Which coins have unusual whale activity in the last 4 hours?"
- "Are longs or shorts getting liquidated on BTC today?"
- "Show the last 10 whale signals that fired and what price did after each."
- "Compare whale buying vs selling on SOL over the past 24 hours."
Install
Recommended: the hosted server. Nothing to install, and it is always current. Point any streamable-HTTP MCP client at it:
{
"mcpServers": {
"coinlobster": { "url": "https://coinlobster.com/mcp" }
}
}
Claude Code
claude mcp add --transport http coinlobster https://coinlobster.com/mcp
VS Code
code --add-mcp '{"name":"coinlobster","url":"https://coinlobster.com/mcp"}'
Claude web / desktop can add it as a connector instead: https://coinlobster.com/mcp/connector (sign in with a free CoinLobster account). Full setup guide: coinlobster.com/connect
Any shell, no MCP client: the hosted server answers plain JSON-RPC over curl (see How it works).
Local stdio, via npm
If you prefer a local process (or your client only speaks stdio), this package is a thin proxy to the same hosted server:
claude mcp add coinlobster -- npx -y coinlobster-mcp
{
"mcpServers": {
"coinlobster": {
"command": "npx",
"args": ["-y", "coinlobster-mcp"]
}
}
}
code --add-mcp '{"name":"coinlobster","command":"npx","args":["-y","coinlobster-mcp"]}'
Either path serves the same data and the same tool schemas.
Tools
17 tools are live on production: 16 canonical plus whale_tape, a deprecated alias kept working for existing callers. Prefer whale_trades. Call tools/list before relying on any tool, the set grows and this table is not the source of truth (npm test only checks that tools/list is non-trivial and that whale_radar answers, so it stays green across additions).
| Tool | What it returns |
|---|---|
whale_trades |
Live whale trades across 15 CEX venues + ETH/Base/Arbitrum DEX swaps, newest first. Filter by coin, source, min USD. |
whale_radar |
Coins with unusual whale flow right now, across 1h / 4h / 24h windows. The best first call. |
flag_outcomes |
Signals that fired, with fire-time price and what price did next. Hits and misses. |
whale_flow |
Hourly whale buy vs sell USD volume for one coin, CEX and DEX split. |
liquidations |
Recent liquidations for a pair with the long / short breakdown. |
market_snapshot |
Price, 24h stats, average funding, open interest and liquidation totals in one call. |
crypto_news |
Recent headlines + a sentiment score (-100 to +100) for a coin, plus the macro picture. |
whale_tape |
Deprecated alias of whale_trades. Same tool, same arguments. |
market_history |
Daily funding, open interest, price and whale-flow history, months deep (30d keyless, full archive on a developer plan, from $29/mo). |
market_extremes |
Pairs at records or their own top/bottom 5% for funding, OI or liquidations, with next-day outcome stats for past occurrences. |
whale_context |
Whale flow vs the coin's own logged history: multiple of its average, N-day highs, buy share (USD magnitudes need depth). |
market_overview |
One-call market scan: totals, 24h whale net flow, BTC risk regime, top rows by volume. |
market_screener |
Screen coins by funding band, OI band, whale multiple or unusual-only. |
funding_matrix |
Per-venue funding rates with predicted next rate where the venue provides it. |
squeeze_score |
Fused 0-100 crowding read with the overexposed side (needs an active developer plan, from $29/mo). |
liq_zones |
Projected liquidation level book, always labeled a projection, never recorded trades (keyless: free pairs, no cascade chains). |
trade_context |
The one-call pre-trade blob: price, funding and OI with percentile judgment vs the pair's own recorded days, whale flow vs baseline, visible liquidations, carry-cost arithmetic. |
All tools are read-only and return descriptive market data only. Never financial advice.
Free tier, and what actually lifts the caps
The free tier needs nothing: BTC/USD and every $1M+ trade stream in real time, other pairs are delayed 15 minutes, DEX swaps $250K and larger, radar magnitudes withheld on non-majors.
A free API key does not lift the MCP caps. The free key is a one-off 1,000-credit grant (coinlobster.com/developers, no card) that unlocks the metered REST endpoints at /api/ai/v1/*, roughly 20 to 100 calls in total. On MCP it changes nothing.
Full MCP depth (live feed on every pair, unlocked radar magnitudes, non-BTC market_snapshot, 30-day flow windows, the full archive) needs an active developer plan: Builder $29/mo (100k credits), Growth $99/mo (500k), Scale $249/mo (2M). One plan covers both full-depth MCP and the monthly REST credits. Present it either as a cl_ key in X-API-Key on the MCP request, or by signing in over the OAuth connector at https://coinlobster.com/mcp/connector with an account that owns a plan.
CoinLobster Pro ($9/mo) is a different product: the coinlobster.com website (live feed, alerts, whale radar). It grants no MCP depth and no API access.
Keyless calls cost nothing. On a developer plan, an MCP tool call costs 2x the mirrored REST endpoint: 40 to 100 credits, by tool (40 for whale_context, liquidations, funding_matrix, liq_zones; 60 for whale_trades, whale_tape, whale_radar, whale_flow, market_history, market_extremes, market_snapshot, market_overview, squeeze_score; 100 for flag_outcomes, crypto_news, market_screener, trade_context). MCP and REST draw on one shared account balance: one plan, one balance, either surface. Discovery (initialize, tools/list, ping) is free, and failed calls and no-data answers are refunded. One-off credit packs top up that same shared balance: $10 = 5,000, $49 = 50,000, $199 = 250,000.
Once the npm package ships, a developer plan key travels in the stdio config:
{
"mcpServers": {
"coinlobster": {
"command": "npx",
"args": ["-y", "coinlobster-mcp"],
"env": { "COINLOBSTER_API_KEY": "cl_your_key" }
}
}
}
How it works
This package is a thin stdio proxy to CoinLobster's hosted MCP server (https://coinlobster.com/mcp, streamable HTTP). Tool schemas and results come from the hosted server verbatim, so new tools appear without a package update. The heavy lifting (exchange streaming, on-chain indexing, signal scoring) runs on CoinLobster's pipeline; your machine just relays JSON-RPC.
Raw call, no client needed:
curl -s -X POST https://coinlobster.com/mcp \
-H 'Content-Type: application/json' -H 'Accept: application/json, text/event-stream' \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"whale_radar","arguments":{}}}'
- Live data pages: whale feed · radar · whale statistics
- Agent-readable docs: llms.txt · crypto-whales.md · radar.md
- REST API: coinlobster.com/developers · OpenAPI: openapi.json
Development
From a local checkout:
npm install
npm test # live smoke: initialize + tools/list + a whale_radar call
MIT licensed. Issues and PRs welcome once the repo is public.
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。
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。
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。
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。
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。
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。
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。
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。