MCP YouTube Transcript
An MCP server retrieving transcripts of YouTube videos
作者提供的说明(英文原文)
Crypto intelligence for AI agents. Hyperliquid trader analytics, cohorts, and live market data.
这三项永远不会合并成一个分数:意见、测量和人气是不同的东西。
我们还没有为它写通俗的介绍。下面是我们从注册中心读取到的事实。
选择你使用的 AI 应用,然后复制设置。
COINVERSAA_API_KEY标为机密的值是你账户的钥匙。千万不要分享给别人,也不要粘贴到聊天里。
你需要
你用的是哪个应用?
在上方选择你的应用。我们会记住它,下次按钮会显示“安装到 …”。
打开 Claude Desktop → Settings → Developer → Edit Config,把下面的内容粘贴到 claude_desktop_config.json
{
"mcpServers": {
"coinversaa": {
"command": "npx",
"args": [
"-y",
"@coinversaa/mcp-server"
],
"env": {
"COINVERSAA_API_KEY": "your-value"
}
}
}
}在终端中运行这条命令
claude mcp add coinversaa --env COINVERSAA_API_KEY=your-value -- npx -y @coinversaa/mcp-server运行前请先填写空着的值。
粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"coinversaa": {
"command": "npx",
"args": [
"-y",
"@coinversaa/mcp-server"
],
"env": {
"COINVERSAA_API_KEY": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
粘贴到项目中的 .vscode/mcp.json
{
"servers": {
"coinversaa": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@coinversaa/mcp-server"
],
"env": {
"COINVERSAA_API_KEY": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
这个服务器在你的电脑上运行,而 ChatGPT 只能连接互联网上的服务器。请选择其他应用。
你的选择通过 Cookie 保存在这台设备上,无需账户。
这些设置是根据作者在注册中心的条目生成的。如果 README 的写法不同,请以 README 为准。
它在你的电脑上运行;安装后,工具列表会显示在你的 AI 应用中。
我们检查的是任何人都能看到的信息;我们没有阅读代码。不适用的检查项不计入总数。 · 检查于 2026年11月6日
注册名称已通过 GitHub 账户 coinversaa 验证,该账户拥有此仓库
我们还没有从 GitHub 读取这个仓库
我们还没有从 GitHub 读取这个仓库
它在你的电脑上运行,没有互联网端点
在你的电脑上运行,改用本地设置
它在你的电脑上运行;我们只能读取互联网端点的工具列表
与注册中心里这个分类中所有活跃的服务器比较。只有当发布者拥有该仓库时,星标数才计入。
它在 1 项适用的基础检查中通过了 1 项。我们没有阅读代码。在添加密钥或令牌之前,请先看看这个服务器要求哪些权限,并使用只具备所需权限的密钥。
你需要填写:COINVERSAA_API_KEY
它在你的电脑上运行。你的 AI 应用会在需要时启动它,数据不会经过作者的服务器(除非它自己调用在线服务)。
GitHub 仓库最后更新于–,最新版本(0.2.1)于8个月前进入注册中心。
支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。你的电脑上还需要有 Node.js。
同一分类中有:MCP YouTube Transcript, Google Analytics MCP, BibiGPT YouTube Video Summarizer
同一分类中其他作者的服务器,按星标数排序。
如何看卡片:★ GitHub 星标i三大支柱:评价 · 检查 · 使用情况i更新 = 最新版本发布到注册中心的时间i
An MCP server retrieving transcripts of YouTube videos
GA4 MCP server for AI agents: schema discovery, server-side aggregation, analysis-ready defaults.
AI video summarizer: summarize and transcribe YouTube, Bilibili, TikTok and podcast videos.
按原文从作者的仓库复制。PopNx MCP 没有审查或背书其内容。HTML、图片和脚本已被移除。链接会离开本站。
Crypto intelligence for AI agents. Query the full Hyperliquid wallet universe, indexed trade history with PnL attribution, behavioral cohorts, and live market data through any MCP-compatible client. Call pulse_global_stats to see exact current coverage (tracked wallets, indexed trades, volume, PnL, data window).
Now with HIP-4 outcome contracts and builder dex support — inspect prediction-market style outcome contracts, settlements, commodities (gold, silver, oil), stocks (TSLA, AAPL), and perps across 8 dexes and 369+ markets.
L4 order book tools and dataset coverage. Five more tools, and full parity with the hosted connector at https://mcp.coinversa.ai/mcp.
| New tool | What it answers |
|---|---|
book_summary [Pro] | "What does the resting book for coin X look like right now — depth, spread, imbalance?" |
book_levels [Pro] | "Where is the resting size stacked, level by level?" |
book_whales [Pro] | "Which wallets are holding the largest resting orders, and on which side?" |
book_stop_map [Pro] | "Where are the trigger orders clustered that a move would sweep?" |
data_coverage | "What window does each dataset cover, and how fresh is it?" |
Tool count: 103 → 107 registered, 106 advertised (see below).
Also in 0.12.0:
COINVERSAA_HIDDEN_TOOLS= (explicitly empty) to advertise everything, or name a different comma-separated list to withhold other tools. Leaving the variable unset withholds builder_heatmap.list_markets, which is the canonical market-discovery tool.live_mark_dislocations, pulse_cohort_bias_history, pulse_cohort_performance_daily, hip4_outcomes, hip4_trader_outcomes and pulse_lifecycle gained the paging and filter parameters the hosted connector already had, and live_coin_risk_history gained include.Builder user-lifecycle, journey, heatmap, and order-intent analytics. 4 more tools over the same builder-fee data, covering where a builder's users stand today, how fast it monetizes a new wallet, when its flow actually trades, and what its users intend at order placement.
| New tool | What it answers |
|---|---|
builder_journey [Pro] | "How fast and how unevenly does builder X monetize a new user?" |
builder_lifecycle [Pro] | "How many of builder X's users are still active, and how many did a rival take?" |
builder_heatmap [Pro] | "What hours does builder X's volume peak, and when is it safe to ship?" |
builder_orders [Pro] | "Do builder X's users place stops and take-profits, and how much of their order flow actually fills?" |
Tool count: 99 → 103.
Also in 0.11.1:
title and MCP annotations (readOnlyHint: true, destructiveHint: false, idempotentHint: true, openWorldHint: true), so clients can label the tools and treat them as read-only without prompting.https://mcp.coinversa.ai/mcp. The obsolete header-authenticated HTTP entrypoint that used to live in this repo has been removed; see Quick Start for the two supported ways to connect.Builder analytics — 8 new tools (91 → 99). The Hyperliquid builder-code economy (the fees frontends, wallet apps, bots, and builder dexes earn on routed order flow) is now a first-class tool family:
builder_leaderboard — builders ranked by exact on-chain ledger revenue, with attributed volume/users and prev-window deltas *(Starter)*builder_profile — one builder: revenue, daily series, top coins, profitable-user share *(Starter)*builder_traders — wallets trading via a builder, with exchange-wide cohort tiers *(Pro)*builder_fills — individual attributed fills through a builder, perp/spot/HIP-4 *(Pro)*builder_cohorts — a builder's user base split by behavioral tier *(Pro)*builder_retention — monthly new-user retention triangle *(Pro)*builder_overlap — which other builders share this one's users *(Pro)*trader_builders — every builder one wallet trades through, by fees paid *(Starter)*Revenue is computed from Hyperliquid's own cumulative builder-fee ledger (reconciled on-chain); fill-level detail comes from order→fill attribution with honest coverage caveats in every response (dataNotes). Ask your agent *"which frontends do MetaMask's traders also use?"* — no other data source can answer that.
Dual-vocabulary cohort tiers. New canonical tier slugs accepted everywhere a tier is an input — PnL tiers (apex, sharps, grinders, scrapers, crowd, bleeders, trapped, blown_out) and weight-class size tiers (heavyweights … strawweights). Legacy slugs (money_printer, smart_money, …) remain valid indefinitely; responses are unchanged. No breaking changes.
Entity resolution, tier-aware sessions, and chain-verified answers.
| New tool | What it answers |
|---|---|
pulse_entity_profile [Pro] | "Who owns this wallet — and what is their REAL combined book across every sub-account?" |
pulse_entity_leaderboard [Pro] | "Top traders deduped by OWNER, not wallet — a fund running 35 sub-accounts shows as one entity." |
pulse_my_plan | "What plan is this API key on, what are the limits, and what does upgrading unlock?" |
pulse_exchange_volume | "24h volume — total and per dex (builder dexes are ~43% and most trackers miss them)." |
pulse_exchange_oi | "Open interest by dex with long/short split." |
pulse_active_traders | "How many wallets traded in the last 24h?" |
pulse_exchange_positions | "How many positions are open right now, per dex?" |
pulse_pnl_leaders | "Who made and lost the most in the last 24h, exchange-wide?" |
Also in 0.9.0:
Tool count: 83 → 91.
Position lifecycles, execution quality, and trader archetypes. v0.8.0 adds 28 tools built on a fully re-derived position-lifecycle dataset — every open→close cycle reconstructed from on-chain fills, now carrying MAE/MFE (the worst adverse and best favorable price each position ever saw). This unlocks execution-quality analysis, not just PnL.
| New tool | What it answers |
|---|---|
pulse_trader_lifecycles | "Show me every open→close position for this wallet, with entry/exit and hold time." |
pulse_trader_lifecycle_summary | "What are this wallet's position-level stats — win rate, avg hold, biggest win/loss?" |
pulse_lifecycle | "Break down lifecycle 12345 into every fill that built and unwound it." |
pulse_trader_demo | "Give me a quick wallet brief before a deeper dive." |
pulse_wallet_drawdown_curve | "How far underwater did each of this wallet's positions go before working?" |
pulse_max_pain_events | "Which winners survived the deepest drawdowns before recovering?" |
pulse_perfect_exits | "Which exits captured most of the maximum favorable move?" |
pulse_backstop_events | "What were the most catastrophic individual liquidations?" |
pulse_survivors / pulse_anti_survivors | "Who blew up and recovered — and who never did?" |
pulse_persistent_winners | "Who is profitable across multiple distinct months, not just lucky once?" |
pulse_capital_titans | "Who extracts the most PnL per dollar of fees paid?" |
pulse_one_month_wonders | "Who had one huge month then gave it back?" |
pulse_newcomer_whales | "Who just showed up and is already trading big notional?" |
pulse_coin_kings | "Who is the top earner of each coin?" |
pulse_top_liquidators | "Who profits most by liquidating others?" |
pulse_lethal_coins | "Which coins blow people up most often?" |
pulse_coin_alpha_map | "Per coin, how big are the winner vs loser profit pools?" |
pulse_hour_profitability | "What UTC hour of close is most profitable?" |
pulse_market_concentration | "How concentrated is alpha — do the top 1% take everything?" |
pulse_style_distribution | "Do scalpers or swing traders make more money?" |
pulse_compare | "Head-to-head: who is the better trader, A or B?" |
pulse_cohort_recent_* | "What are wallets that are printing RIGHT NOW (last-30-day tier) doing — positions, trades, top lifecycles, concentration?" |
pulse_lifecycles_recent | "What just closed exchange-wide right now?" (global feed; successor to pulse_recent_closed_positions) |
The legacy closed-position tools (pulse_trader_closed_positions, pulse_trader_closed_position_stats, pulse_recent_closed_positions) are kept for backward compatibility but superseded by the lifecycle tools, which read the corrected position_lifecycles_full table (more history, MAE/MFE, spot).
Tool count: 55 → 83. An API key is required for every tool; backend tiering determines which tools and limits are available.
HIP-4 outcome contract intelligence. v0.7.0 adds 12 tools for discovering active outcomes, reading question metadata, inspecting settlements and recent fills, tracking daily volume, ranking outcome traders, measuring outcome/perp overlap, and joining outcome holders to their currently open perp positions on the same underlying asset.
| New tool | What it answers |
|---|---|
hip4_outcomes | "What outcome contracts are active right now?" |
hip4_outcome | "What is outcome 123 and what side tokens does it use?" |
hip4_outcome_summary | "How much volume/PnL has this outcome done across both sides?" |
hip4_outcome_recent_trades | "Show me recent fills for this prediction market." |
hip4_questions | "What HIP-4 questions and named outcomes exist?" |
hip4_recent_settlements | "Which outcomes settled recently and which side won?" |
hip4_daily_volume | "Is HIP-4 outcome volume growing day by day?" |
hip4_most_active | "Which outcome contracts are most active?" |
hip4_top_traders | "Who are the top outcome traders?" |
hip4_trader_outcomes | "What outcomes did this wallet trade?" |
hip4_cross_product_overlap | "How much overlap is there between outcome traders and perp traders?" |
hip4_perp_position_context | "Do outcome 25 traders currently have open BTC perp exposure, and is it aligned or hedged?" |
Tool count: 43 → 55. An API key is required for every tool; backend tiering determines which tools and limits are available. Get a key from coinversa.ai/developers.
Canonical cross-market asset taxonomy. The same underlying asset can appear under different tickers on different venues (e.g. GOLD on xyz, PAXG on native Hyperliquid — both track gold). v0.6.0 added 3 tools that resolve synonyms server-side and aggregate across venues, plus a ground-truth OI tool:
| New tool | What it answers |
|---|
| list_assets | "What assets
只显示了开头部分;其余内容请在仓库中阅读