MCP Client (Claude Desktop / Cursor / etc.)
            │
            ▼
┌───────────────────────────────────┐
│    EnterpriseMasterGateway        │  ← server.py
│    FastMCP Router                 │
│                                   │
│   namespace: "market"             │   namespace: "quant"
│   ┌───────────────────────┐       │   ┌───────────────────────┐
│   │  CryptoData Server    │       │   │  CryptoQuant Server   │
│   │  data_server.py       │       │   │  quant_server.py      │
│   │                       │       │   │                       │
│   │  Resource:            │       │   │  Tool (app=True):     │
│   │  data://trending      │       │   │  quant_pipeline       │
│   │                       │       │   │  → Gemini 2.5 Flash   │
│   │  Tools:               │       │   │  → MarketSignal       │
│   │  get_crypto_price     │       │   │    (Pydantic)         │
│   │  query_and_filter     │       │   │  → PrefabApp UI       │
│   └───────────────────────┘       │   └───────────────────────┘
└───────────────────────────────────┘
            │
            ▼
    otel_launcher.py
    OpenTelemetry Auto-Instrumentation
    (OTLP export if configured, silent otherwise)

enterprise-crypto-mcp/
│
├── server.py           # Master gateway — mounts both sub-servers with namespacing
├── data_server.py      # CryptoData sub-server — prices, trending, filtering
├── quant_server.py     # CryptoQuant sub-server — AI quant pipeline + UI rendering
├── otel_launcher.py    # OpenTelemetry configuration and auto-instrumentation launcher
│
├── .env                # Environment variables (gitignored)
├── .gitignore
└── requirements.txt    

---

## Sub-Servers — Detailed Breakdown

### `CryptoData` — `data_server.py`

Handles all raw market data retrieval from the CoinGecko public API. Exposes one MCP Resource and two Tools.

---

####  Resource — `data://trending`

Returns a formatted snapshot of the top 5 trending cryptocurrencies on CoinGecko, including coin ID, ticker symbol, and market cap rank.

| Property      | Value              |
|---------------|--------------------|
| Source API    | CoinGecko `/search/trending` |
| Cache TTL     | **300 seconds** (5 minutes) |
| Cache strategy| In-memory dict with expiry timestamp |
| OTel span     | `get_trending_markets_resource` with `cache.status` attribute |

---

####  Tool — `get_crypto_price`

Fetches the current USD price of a single token by CoinGecko coin ID (e.g. `bitcoin`, `ethereum`). Returns a clean `float` value optimized for downstream arithmetic operations without any formatting overhead.

| Property      | Value              |
|---------------|--------------------|
| Source API    | CoinGecko `/simple/price` |
| Cache TTL     | **60 seconds** (1 minute) |
| OTel span     | `get_crypto_price_execution` with `target.coin` and `cache.status` attributes |

---

####  Tool — `query_and_filter_market_data`

Accepts a comma-separated list of coin IDs and a USD threshold. Evaluates all price lookups **locally in code** and passes only coins that cross the specified threshold back to the LLM — discarding the rest entirely.

This is a deliberate **context window optimization pattern**: for large watchlists, sending every token's price to the LLM is wasteful. By filtering on the server side, only the signal-bearing subset reaches the model.

| Property         | Value |
|------------------|-------|
| Input            | `"bitcoin,ethereum,solana,cardano"`, threshold `50.0` |
| Output           | Only coins with price ≥ threshold, formatted as `COIN: $price USD` |
| OTel span        | `code_api_data_filtering` with `filter.input_count` and `filter.output_count` attributes |

---

### `CryptoQuant` — `quant_server.py`

Runs a full AI-backed quantitative analysis pipeline on a set of target tokens. Returns a structured `PrefabApp` UI object rendered directly in the MCP client.

---

####  Tool (UI) — `advanced_crypto_quant_pipeline`

**Pipeline stages:**

| Stage | Action |
|-------|--------|
| 1 | Fetches live prices for all target coins via `get_crypto_price` (reuses CryptoData tool directly, preserving its internal telemetry spans) |
| 2 | Assembles a price matrix string and applies optional risk mitigation flag |
| 3 | Calls **Gemini 2.5 Flash** with `responseMimeType: application/json` to generate a structured market signal |
| 4 | Validates the raw JSON response against the `MarketSignal` Pydantic model |
| 5 | Builds and returns a `PrefabApp` UI layout using `prefab_ui` components |

Progress is reported at each stage via `ctx.report_progress(current, total, message)`, giving MCP clients real-time execution feedback.

**`MarketSignal` Schema (Pydantic):**

```python
class MarketSignal(BaseModel):
    sentiment: str        # "Bullish" | "Bearish" | "Neutral"
    entry_target: str     # Optimized entry zone description
    stop_loss: str        # Defensive stop condition description
    risk_score: int       # 1–10 integer risk rating
    synthesis: str        # Exactly one sentence of analytical summary

git clone https://github.com/your-username/enterprise-crypto-mcp.git
cd enterprise-crypto-mcp

python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate

pip install fastmcp httpx python-dotenv opentelemetry-distro prefab-ui pydantic

# Required
GEMINI_API_KEY=your_google_ai_studio_key_here

# Optional — OpenTelemetry export (omit to disable silently)
OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317
OTEL_SERVICE_NAME=EnterpriseMasterGateway

python server.py

python otel_launcher.py

{
  "mcpServers": {
    "enterprise-crypto": {
      "command": "python",
      "args": ["/absolute/path/to/otel_launcher.py"],
      "env": {
        "GEMINI_API_KEY": "your_key_here"
      }
    }
  }
}