[ TECHNICAL DOCUMENTATION ]

MCPTRADE
DEVELOPER REFERENCE

Complete technical reference for the MCPTrade intelligence platform — MCP protocol integration, data pipeline architecture, skill library API, and database schema.

[ Overview ]

MCPTrade is a Next.js 16 application that uses Claude AI connected to TradingView Desktop via the Model Context Protocol (MCP) to perform automated technical analysis on the top 100 cryptocurrencies. Analysis runs offline via a Python script and results are stored in Supabase. The frontend is a read-only SSR dashboard that revalidates every hour.

Runtime

Next.js 16 / Turbopack

AI Model

Claude Sonnet 4.5

Database

Supabase (Postgres)

MCP Server

tradingview-mcp

Data Source

CoinGecko API v3

Revalidation

ISR · 3600s

[ Architecture ]

The system separates data generation from data presentation. No analysis happens at request time. The Python script runs as a one-shot batch job and writes results to Supabase. The Next.js frontend reads from Supabase at build time and caches pages with ISR.

Generation Layer— offline · run manually or via cron

generate_analysis.py

Entry point

CoinGecko API

OHLCV · Top 100

Anthropic API

Claude Sonnet · signal

1.fetch_top100_coingecko() → coin list (id, name, price)

2.fetch_ohlcv_coingecko() → 365-day OHLCV per coin

3.detect_support_resistance() + calc_rsi / calc_macd / calc_ema

4.claude.messages.create() → { signal, score, narasi }

5.cache_to_supabase() → upsert coin_analysis

6.[ optional ] mcp_connect_tradingview() → screenshot_chart() → Claude vision

Supabase Postgres · upsert coin_analysis

Presentation Layer— Next.js 16 · Vercel · ISR revalidate=3600

Dashboard

SSR · CoinTable

/[coinId]

Coin Detail

Chart · Terminal · S/R

/skills

Skills

58 SkillCards

/skills/[id]

Skill Detail

Code · About · Related

/docs

Docs

Static reference

[ MCP Protocol ]

The Model Context Protocol connects Claude to external tool servers via a standardized JSON-RPC transport. MCPTrade uses the tradingview-mcp server which exposes TradingView Desktop as a set of callable tools.

Available MCP Tools

Tool Name

Description

connectEstablish WebSocket connection to TradingView MCP server

set_symbolNavigate chart to a specific symbol and timeframe

get_chart_dataRead OHLCV bar data from the active chart

draw_lineDraw a horizontal line at a specified price level

add_studyAdd a built-in TradingView indicator to the chart

screenshotCapture chart as PNG, returned as base64 bytes

create_alertCreate a price alert with specified condition and notification methods

get_watchlistReturn all symbols in the active TradingView watchlist

set_timeframeChange the active chart timeframe

python — MCP connection pattern

import anthropic_mcp as mcp

# 1. Connect once per session
client = mcp.MCPClient(server="tradingview-mcp", timeout=30)
conn = client.connect()
assert conn.status == "connected"

# 2. Navigate and read
client.call("set_symbol", {"symbol": "BINANCE:BTCUSDT", "timeframe": "1D"})
client.wait_for("chart_loaded")
data = client.call("get_chart_data", {"bars": 365, "timeframe": "1D"})

# 3. Analyze and draw
levels = detect_sr(data["high"], data["low"])
for price in levels["support"]:
    client.call("draw_line", {"price": price, "color": "#22C55E", "style": "solid"})

[ Data Flow ]

Analysis is a three-phase pipeline: data ingestion from CoinGecko, computation via Python indicator functions and Claude API, then persistence to Supabase. The frontend never calls CoinGecko or Anthropic — it reads exclusively from Supabase.

sequence


  generate_analysis.py              CoinGecko API          Supabase
       │                                 │                      │
       │── fetch_top100() ──────────────>│                      │
       │<── [ {coin_id, name, price} ] ──│                      │
       │                                 │                      │
       │── fetch_ohlcv(coin_id) ────────>│                      │
       │<── [ [ts, O, H, L, C] × 365 ] ──│                      │
       │                                 │                      │
       │  calc_rsi / calc_macd / detect_sr  (local)             │
       │  claude.messages.create()          (Anthropic API)     │
       │                                                        │
       │── sb.table("coin_analysis").upsert(result) ──────────> │
       │<── { status: "ok" } ──────────────────────────────────│
                                                                │
  Next.js DashboardPage ─── getAllCoins() ─────────────────────>│
                         <── CoinAnalysis[] ────────────────────│

[ Quick Start ]

bash

# 1. Clone and install
git clone <repo-url> && cd mcptrade
pnpm install

# 2. Set environment variables (see Environment section)
cp .env.example .env.local

# 3. Populate the database (one-time or on refresh)
cd scripts
python generate_analysis.py

# 4. Start the development server
cd ..
pnpm dev

Prerequisites

Node.js≥ 20.xRequired for Next.js 16 and pnpm

Python≥ 3.10Required for generate_analysis.py

TradingView DesktopLatestRequired for MCP chart tools (optional for data-only mode)

pnpm≥ 9.xPackage manager (npm/yarn also work)

[ Environment ]

Variable

Scope

Purpose

NEXT_PUBLIC_SUPABASE_URLclientSupabase project URL — exposed to browser

NEXT_PUBLIC_SUPABASE_ANON_KEYclientSupabase anon key — read-only access, safe to expose

ANTHROPIC_API_KEYserverAnthropic API key — used by generate_analysis.py only

SUPABASE_SERVICE_KEYserverSupabase service role key — used for upsert operations in script

.env.local

NEXT_PUBLIC_SUPABASE_URL=https://your-project.supabase.co
NEXT_PUBLIC_SUPABASE_ANON_KEY=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9...
ANTHROPIC_API_KEY=sk-ant-...                     # script only
SUPABASE_SERVICE_KEY=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9...  # script only

[ Skills Reference ]

58 Python functions organized into 7 categories. Each skill is self-contained with no cross-skill dependencies unless explicitly noted. All functions assume Python 3.10+.

MCP Integration

Connect, control, and read data from TradingView Desktop via the MCP protocol

Chart Analysis

Detect S/R, trends, patterns, divergence, and breakouts from OHLCV arrays

Market Scanner

Scan coin lists for specific signal setups — buy signals, crosses, extremes

Indicators

Pure Python RSI, EMA, MACD, ATR, VWAP, Ichimoku — no NumPy/pandas required

Pine Script

Ready-to-paste Pine Script v5 indicators and strategies for TradingView

Data Pipeline

Fetch, normalize, batch, schedule, and export coin analysis data

Claude AI

Prompt templates for narrative generation, scoring, and visual chart analysis

VIEW ALL 58 SKILLS →

[ API Reference ]

GET /api/debug

Returns the raw Supabase connection status and a sample of the first coin record. Useful for verifying that environment variables are correctly configured after deployment.

lib/supabase.ts — getAllCoins()

returnCoinAnalysis[]All rows from coin_analysis ordered by market_cap_rank ascending

revalidate3600sISR — pages using this function cache for 1 hour, then revalidate on next request

throwsneverErrors are caught internally; empty array returned when Supabase is unavailable

lib/supabase.ts — getCoin(coinId)

coinIdstringCoinGecko coin ID (e.g. 'bitcoin', 'ethereum')

returnCoinAnalysis | nullSingle coin record or null if not found

[ Database Schema ]

Table: coin_analysis

Column

Type

Description

coin_idtext✓CoinGecko ID — primary key, conflict target for upsert

nametextDisplay name (e.g. "Bitcoin")

symboltextTicker symbol lowercase (e.g. "btc")

tv_symboltextTradingView symbol string (e.g. "BINANCE:BTCUSDT")

current_pricefloat8Last known USD price from CoinGecko

price_change_24hfloat824-hour price change as a percentage

market_cap_rankint4Market cap ranking at time of analysis

image_urltextCoinGecko CDN URL for the coin logo PNG

signaltextClaude's signal output: "buy" | "sell" | "neutral"

scoreint4Signal quality score 0–100 from rate_signal_quality()

trendtextTrend classification: "Bullish" | "Bearish" | "Sideways"

rsifloat814-period RSI at time of analysis

macdtextMACD state string (e.g. "Positive" | "Negative")

ema_crosstextEMA cross state (e.g. "Golden Cross" | "Death Cross")

supportjsonbArray of {price, strength} support level objects

resistancejsonbArray of {price, strength} resistance level objects

narasitextClaude-generated 2-sentence market narrative

analyzed_attimestamptzUTC timestamp when the analysis was generated

sql — create table

create table coin_analysis (
  coin_id         text primary key,
  name            text,
  symbol          text,
  tv_symbol       text,
  current_price   float8,
  price_change_24h float8,
  market_cap_rank  int4,
  image_url       text,
  signal          text,
  score           int4,
  trend           text,
  rsi             float8,
  macd            text,
  ema_cross       text,
  support         jsonb,
  resistance      jsonb,
  narasi          text,
  analyzed_at     timestamptz default now()
);

MCPTRADEDEVELOPER REFERENCE

MCPTRADE
DEVELOPER REFERENCE