Followin Skill User Guide

English | 简体中文

13 AI agent skills for crypto trading, macro analysis, and market intelligence — powered by Followin MCP and Premium MCP.

All skills support both Chinese and English triggers and output in the language you use.

What's portable across AI tools

The package has two layers:

1. MCP servers (Followin MCP + Premium MCP) — these provide the underlying data and tools, and work in any MCP-compatible AI client: Claude Code, Claude Desktop, Cursor, Windsurf, Cline, Continue.dev, OpenCode, OpenClaw, and so on.
2. Skill files (the 13 markdown files in .claude/commands/) — written in Claude Code's slash-command format. They drop in directly to Claude Code and other clients that read the same format. For tools that use a different rule/command format, the markdown bodies are still fully portable — just paste the relevant fields into your tool's native format.

Setup

npx @followin/skills setup

Paste your Followin API key when prompted, then restart your client. Done — the 13 skill files are installed and both MCP servers (followin-mcp + premium-mcp) are configured and validated.

Default layout (1.6.0+): skills project-local, MCP global. npx @followin/skills setup writes skill files to <cwd>/.claude/commands/ (only active in this project) and MCP config to ~/.claude.json (available in every Claude Code session). Configure your API key once, use the MCP servers everywhere. For a fully global install (skills too), pass --client claude-code.

Other clients:

npx @followin/skills setup --client claude-code     # all Claude Code projects (global)
npx @followin/skills setup --client cursor          # run from your project dir
npx @followin/skills setup --client windsurf        # run from your project dir
npx @followin/skills setup --client claude-desktop  # MCP only
npx @followin/skills setup --client opencode-project # OpenCode / OpenClaw (skills local, MCP global)
npx @followin/skills setup --client opencode        # OpenCode / OpenClaw (global)

Cursor and Windsurf get the rules auto-converted into their native formats (.cursor/rules/*.mdc and .windsurf/rules/*.md) — run setup from inside the project directory where you want them. Claude Desktop is MCP-only because there's no stable filesystem path for skills there yet.

Run npx @followin/skills clients to see every preset. Requires Node.js 16+ (macOS / Linux / Windows). The API key is stored in plaintext in the client config (chmod 600) — don't commit that file.

{
  "mcpServers": {
    "followin-mcp": {
      "type": "sse",
      "url": "https://mcp.followin.io/sse?api_key=YOUR_API_KEY_HERE",
      "headers": { "X-API-Key": "YOUR_API_KEY_HERE" }
    },
    "premium-mcp": {
      "type": "sse",
      "url": "https://premium-mcp.followin.io/sse?api_key=YOUR_API_KEY_HERE",
      "headers": { "X-API-Key": "YOUR_API_KEY_HERE" }
    }
  }
}

Followin MCP — Crypto News & Sentiment

Followin MCP connects to the Followin crypto intelligence platform, providing:

• Trending Topics — Top 10 hot topics with heat scores, related tokens, and price movements
• Trending Feed — Popular news articles and tweets ranked by engagement
• Token-level Content — Key events, news, and community buzz filtered by specific token
• Token KOL Opinions — Twitter KOL views and analysis on specific tokens
• Daily Digest — Curated daily content from multiple media/KOL/institutional sources
• Keyword Search — Search articles, news flashes, and tweets with type filtering and time range
• Intel Center — Structured data channels: token unlocks, macro data, listing/delisting, project events, volume anomalies, funding rates

Built on these capabilities, 5 skills:

01 Followin Intel Center

One-stop query for market event data: token unlocks, macro economic releases, exchange listing/delisting announcements, project events, volume anomalies, and funding rates. Filter precisely by the dimension you care about.

How to ask:

• Any large token unlocks this week
• What coins got listed recently
• Which tokens have abnormal funding rates
• Intel center overview — top 1-2 items from each of the 6 channels

6 data channels:

02 Breaking News Analysis

Paste a crypto news headline or article, get back which tokens are affected, whether it's bullish or bearish, how strong the impact is, and for how long. No MCP dependency — pure AI analysis with optional web search verification.

How to ask:

• Paste a news snippet + Analyze this news
• Analyze news impact
• Is this bullish or bearish

Two modes:

• Flash mode — short news (1-3 sentences): quick bull/bear call with key targets
• Deep mode — long article or explicit request: full causal chain, second-order effects, historical parallels

Every bull/bear call includes a counter-argument — what conditions would invalidate the thesis.

03 Trending News & Topics

What is the market paying attention to right now? Cross-references the trending topic rankings with trending articles/tweets to find multi-platform resonance — topics that appear in both sources get top priority.

How to ask:

• What's hot today
• Trending news today
• Hot topics today
• What's the market focused on

Two modes:

• Overview (default) — full landscape of current hot events
• Focused — specify a token or topic: Any news on ETH, What's happening in the AI sector

Events are ranked by importance (macro/regulatory > sector-wide > single project), with resonance topics always on top.

04 Crypto Daily Brief

Generates a ready-to-publish daily crypto market briefing. Pulls from trending topics + daily digest sources, filters past 12 hours, selects 8-10 most important events, cross-validates, and outputs in a punchy editorial style.

How to ask:

• Crypto daily
• Crypto morning brief
• Give me a daily brief

Key rules:

• Item #1 is always BTC price action in editorial tone (not cold data)
• At least 2-3 macro/AI items guaranteed (won't be crowded out by altcoin noise)
• ~1/3 of items include a one-line editorial comment
• Altcoins outside top 10 by market cap limited to 2-3 items max

Want macro/US stock morning brief instead? Say morning brief or US stock daily — that routes to Skill 10.

05 Token Buzz & Views

Specify a token, get a 4-dimension snapshot: news flashes, in-depth articles, Twitter KOL opinions, and community discussions — all in one query.

How to ask:

• Any news on BTC
• What's happening with ETH
• SOL community discussion — single dimension only
• HYPE articles

Want KOL trading strategy views on a token? Say What are KOL calls on BTC — that routes to Skill 06 (Trading Strategy).

Premium MCP — Trading Strategy & Macro Analysis

Premium MCP aggregates professional data sources:

• Top Trader Positions — Real-time CEX contract trader positions (direction, leverage, size)
• On-chain Whales — Hyperliquid on-chain verifiable whale positions (entry price, liquidation price, exact size)
• KOL Call Orders — KOL publicly shared trading strategies (entry/target/stop)
• TG Channel Aggregation — 70+ Telegram channels across 10 categories
• Crypto Real-time Prices — Batch USD prices for crypto assets
• FRED Macro Data — Federal Reserve economic database (rates, inflation, employment, GDP, yields)
• Financial Market Data — US stock/ETF/index/forex/commodity quotes, financial statements, earnings Beat/Miss, analyst estimates, key ratios, insider trading, economic calendar, treasury yields, gainers/losers
• 31 Financial Media Search — Keyword search across Reuters, Bloomberg, CNBC, WSJ, FT and more
• Twitter Data — User tweets, search, follow relationships, Spaces

Built on these capabilities, 8 skills:

06 Trading Strategy & Signal

Three data sources cross-validated: CEX trader real positions + Hyperliquid on-chain whales + KOL public calls. Core principle: money talks louder than tweets — real positions always outweigh verbal calls.

How to ask:

• Can I trade SOL — full 3-source cross-validation
• What are traders doing — overview of all active positions
• What are KOL calls on ETH — KOL calls only
• Whale positions — on-chain positions only
• Who's longing / Who's shorting

Signal classification:

07 TG Channel Intel

Extracts intelligence from 70+ Telegram channels, clusters by topic, distills views, identifies consensus and disagreements. Channels are quality-graded (high/medium), with high-quality channel views shown first. Supports Chinese, English, Korean, and Vietnamese language channels.

10 channel categories: Macro Analysis, Market Structure, Project Research, Narrative Tracking, Trading Signals, Live Trading, On-chain Data, Meme/New Launches, Cross-market, News Aggregation

How to ask:

• What's TG talking about — full overview
• TG discussion / TG intel
• What are Telegram channels saying about BTC — focused on a token
• What are trading channels saying — filtered by category

Auto-routing by user profile:

08 BTC Macro Dashboard

Quantitative 0-100 macro environment score for BTC based on 15 indicators across 4 layers: global liquidity (35%), market environment (30%), crypto-native capital flows (25%), and economic data pulse (10%). Score of 50 is neutral; above 50 bullish, below 50 bearish.

4-layer scoring:

• Layer 1 — Liquidity Direction (35%): Net liquidity trend, Fed policy, FedWatch probabilities, M2 trend
• Layer 2 — Market Environment (30%): DXY, Nasdaq, VIX, real rates (TIPS), yield curve, gold
• Layer 3 — Crypto-native Capital (25%): BTC spot ETF flows, stablecoin market cap trend
• Layer 4 — Economic Data Pulse (10%): CPI/PCE surprise, employment surprise

Includes contradiction detection — flags when different layers point in opposite directions.

How to ask:

• BTC macro / BTC macro dashboard
• BTC macro score — quick score only
• How is BTC macro

Heavy MCP usage: pulls 10+ FRED series, multiple market quotes, DeFiLlama API, and web searches. Not recommended for high-frequency use.

09 Gold Macro Dashboard

Quantitative 0-100 macro environment score for gold based on 15 indicators across 5 layers: real rates & monetary policy (40%), USD & FX (15%), central bank buying & physical demand (25%), safe-haven demand (15%), and economic data pulse (5%).

Key difference from BTC dashboard:

• Gold's #1 driver is real interest rates (TIPS yield), not liquidity
• VIX scoring is reversed: VIX spike = bullish for gold (safe haven), bearish for BTC (risk-off)
• Central bank buying is a structural factor unique to gold (not in BTC dashboard)
• COT positioning is a contrarian indicator: extreme net longs = bearish (overcrowded)

How to ask:

• Gold macro / Gold macro dashboard
• Gold macro score — quick score only
• How is gold macro

Heavy MCP usage: pulls 8+ FRED series, market quotes, and web searches for FedWatch/COT/ETF/central bank data. Not recommended for high-frequency use.

10 Macro Morning Brief

Daily macro financial morning briefing: full-spectrum treasury yields, VIX, crude oil, USD index, economic calendar, media hot topics, market gainers/losers — one report to grasp the macro landscape before market open.

How to ask:

• Morning brief / Morning briefing
• Macro morning brief
• US stock daily / Macro daily
• Financial morning brief

Report sections:

1. Macro Environment table (10Y/2Y yields, spread, VIX, Brent crude, USD index)
2. Economic Calendar (US events, filtered from full global data)
3. Top 3 Media Hot Topics (with sentiment and representative articles)
4. Watchlist Movers (configurable via watchlist parameter)
5. Market Gainers/Losers (filtered for market cap > $500M)
6. Sentiment Distribution and overall market mood

Heavy MCP usage: 8-9 parallel data pulls + multi-round news searches. Best used once during pre-market hours.

11 US Stock Earnings Report

Three-dimensional single-stock earnings analysis: financial Beat/Miss + media sentiment + macro backdrop. Covers income statement trends, Adjusted vs GAAP EPS, analyst forward estimates, key ratios, price momentum, and 14-day media coverage.

How to ask:

• AAPL earnings / TSLA earnings report
• Show me NVDA earnings
• Earnings analysis MSFT

Key features:

• Beat/Miss uses Adjusted EPS from the earnings endpoint (not GAAP)
• Trend table shows both Adjusted EPS and GAAP EPS side by side
• Forward PEG calculated from analyst estimates
• Price momentum across 7 timeframes (1D to 1Y)
• Media sentiment from 31 financial media sources (Claude-inferred)
• Macro cross-check based on sector (e.g., Tech gets 10Y yield + VIX)

Heavy MCP usage: 7+ parallel data calls per stock. Use during earnings season for quick single-stock reviews.

12 Macro Analyzer

When a macro indicator drops, what sectors and ETFs are affected? Full chain: FRED data trend -> sector ETF actual performance -> media interpretation. Built-in mappings for 20+ common indicators.

How to ask:

• CPI impact / NFP impact
• Rate impact / GDP impact
• Tariff impact
• Macro analysis / Indicator analysis

Built-in indicator mappings:

Two verification modes:

• Real-time (data < 7 days old): ETF moves reflect market's reaction to actual data
• Expectation (data > 7 days old): ETF moves reflect market pricing next release

13 US Stock Divergence Scan

Batch scans the US stock market for inconsistencies between price action, insider trading, and media coverage. Surfaces signals that most investors miss.

4 signal types:

How to ask:

• Divergence scan / US stock divergence scan
• Silent moves / Silent buy
• Anomaly signals
• Unreported drop / Unreported surge

Process: Pulls gainers/losers + insider trading -> filters by market cap -> cross-checks with 31 financial media sources -> flags divergences. Stocks with multiple signals flagged prominently.

Heaviest MCP usage of all skills: requires market cap filtering for each candidate + individual news searches. Not recommended for frequent use.

Quick Reference

Skill Routing Guide

Some requests sound similar but route to different skills. Here's how routing works:

Technical Notes

MCP tool behaviors confirmed through testing:

• finance_tool_quote with ^VIX and ^IXIC — cannot be batched, must call individually
• finance_tool_batch_quote_short — use for DXUSD (quote returns 402), supports comma-separated symbols
• finance_tool_stable_request — required for profile, analyst-estimates, stock-price-change (schema not fixed)
• finance_tool_insider_trading_search — the only working insider trading endpoint; stable_request path insider-trading returns 404
• fred_get_series — limit must be integer (not string); use sort_order: "desc" for latest data first
• search_finance_news — does not support compound keywords; search each concept separately; users must be array type
• finance_tool_economic_calendar — returns ~2.4MB of global data; needs post-processing to filter US events