Pyth Network MCP Server

Official Model Context Protocol (MCP) server for Pyth Network oracle data.

An enterprise-grade TypeScript implementation providing Claude and other AI systems with direct access to Pyth Network's decentralized oracle infrastructure.

---

Overview

This MCP server enables AI agents to interact with Pyth Network's real-time price feeds, historical data, and network information through a standardized protocol. Built for production use with comprehensive error handling, type safety, and performance optimization.

What is Pyth Network?

Pyth Network is a decentralized oracle protocol delivering real-time market data to smart contracts across 107+ blockchains. Backed by 125+ first-party data providers (exchanges, market makers, trading firms), Pyth offers:

• 1,930+ price feeds across crypto, equities, FX, commodities, and rates
• ~400ms update frequency with sub-second latency
• Universal price feed IDs that work across all supported chains
• Confidence intervals for risk-aware applications

---

Features

Tools

Tool	Description
get_price_feeds	Search and filter 1,930+ price feeds by symbol, asset type, or query
get_latest_price	Fetch real-time prices with confidence intervals and publish times
get_price_at_timestamp	Query historical prices at specific Unix timestamps
get_ema_price	Get Exponential Moving Average prices with deviation analysis
get_twap	Calculate time-weighted average prices (1-600 second windows)
get_publisher_caps	Access publisher staking limits and network health data
get_price_feed_info	Detailed metadata for specific price feeds
get_historical_prices	Query historical price data over time ranges
get_ohlcv	Get OHLCV candlestick data (TradingView format)
search_symbols	Search for price feed symbols by name
get_popular_feeds	Curated list of popular feed IDs (BTC, ETH, stablecoins, etc.)

Resources

Resource URI	Description
pyth://network/status	Network health, supported chains, and statistics
pyth://feeds/catalog	Complete catalog of available price feeds
pyth://feeds/popular	Commonly used feeds (BTC, ETH, SOL, major pairs)
pyth://docs/api	Quick reference for Hermes API endpoints
pyth://docs/integration	Integration guides and best practices

Prompts

Prompt	Description
analyze_price_feed	Deep analysis of an asset with price, TWAP, volatility
compare_assets	Side-by-side comparison of multiple price feeds
market_overview	Comprehensive report for an asset category
volatility_report	Historical volatility metrics and risk assessment
price_deviation_check	Detect stablecoin depegs or feed anomalies across USD/USDC/USDT pairs

---

Installation

Prerequisites

• Node.js 18+
• npm or pnpm

Quick Start

# Clone and install
git clone https://github.com/pyth-network/mcp-server.git
cd mcp-server
pnpm install

# Build
pnpm build

# Run
pnpm start

Claude Desktop Configuration

Add to your Claude Desktop config (~/.config/claude/claude_desktop_config.json):

{
  "mcpServers": {
    "pyth-network": {
      "command": "node",
      "args": ["/path/to/pyth-mcp-server/dist/index.js"],
      "env": {
        "PYTH_HERMES_URL": "https://hermes.pyth.network"
      }
    }
  }
}

---

Architecture

pyth-mcp-server/
├── src/
│   ├── index.ts              # Server entry point
│   ├── server.ts             # MCP server implementation
│   ├── api/
│   │   ├── hermes.ts         # Hermes API client
│   │   ├── benchmarks.ts     # Benchmarks API client
│   │   └── types.ts          # API response types
│   ├── tools/
│   │   ├── index.ts          # Tool registry
│   │   ├── price-feeds.ts    # Price feed tools
│   │   ├── historical.ts     # Historical data tools
│   │   └── network.ts        # Network info tools
│   ├── resources/
│   │   ├── index.ts          # Resource registry
│   │   └── handlers.ts       # Resource handlers
│   ├── prompts/
│   │   ├── index.ts          # Prompt registry
│   │   └── templates.ts      # Prompt templates
│   └── types/
│       ├── mcp.ts            # MCP protocol types
│       └── pyth.ts           # Pyth-specific types
├── tests/
├── docs/
└── dist/

---

API Reference

Hermes API (Real-time Data)

Base URL: https://hermes.pyth.network

Endpoint	Purpose
/v2/price_feeds	List all available price feeds
/v2/updates/price/latest	Get latest prices for feed IDs
/v2/updates/price/{timestamp}	Get prices at specific timestamp
/v2/updates/twap/latest	Get TWAP for feed IDs
/v2/updates/publisher_stake_caps/latest	Publisher staking data

Benchmarks API (Historical Data)

Base URL: https://benchmarks.pyth.network

Endpoint	Purpose
/v1/price_feeds/	List price feeds with metadata
/v1/updates/price/{timestamp}	Historical price at timestamp
/v1/updates/price/{timestamp}/{interval}	Price history with intervals
/v1/shims/tradingview/history	TradingView-compatible OHLCV

---

Differences from Community Version

This official implementation provides several advantages over the community-built version:

Feature	Community	Official
Language	Python	TypeScript
Type Safety	Runtime	Compile-time
Streaming	No	SSE support
Historical Data	Basic	Full Benchmarks API
Error Handling	Basic	Comprehensive with retries
Caching	None	Intelligent caching
Rate Limiting	None	Built-in respect for limits
Documentation	Good	Extensive
Maintenance	Community	Pyth Foundation

---

Environment Variables

Variable	Default	Description
PYTH_HERMES_URL	https://hermes.pyth.network	Hermes API endpoint
PYTH_BENCHMARKS_URL	https://benchmarks.pyth.network	Benchmarks API endpoint
PYTH_CACHE_TTL	5000	Cache TTL in milliseconds
PYTH_REQUEST_TIMEOUT	30000	Request timeout in ms
LOG_LEVEL	info	Logging verbosity

---

Development

# Install dependencies
pnpm install

# Run in development mode
pnpm dev

# Run tests (unit tests only)
pnpm test

# Run tests with coverage
pnpm test:coverage

# Run integration tests against live Pyth APIs
RUN_INTEGRATION=true pnpm test

# Build for production
pnpm build

# Lint
pnpm lint

# Type check
pnpm typecheck

Test Suite

The project includes comprehensive tests:

Test File	Tests	Description
types.test.ts	21	Zod schemas, utility functions
errors.test.ts	21	Error classes, wrapping
api.test.ts	19	Mocked API clients
tools.test.ts	18	All 11 tool handlers
resources.test.ts	19	All 5 resources
prompts.test.ts	19	All 5 prompts
server.test.ts	4	Server setup
integration.test.ts	10	Live API tests (skipped by default)

Total: 121 tests

---

Privacy Policy

This MCP server connects to Pyth Network's public APIs:

• Hermes API (https://hermes.pyth.network) - Real-time price data
• Benchmarks API (https://benchmarks.pyth.network) - Historical price data

Data Collection

This server:

• Does NOT collect, store, or transmit any user data
• Does NOT require authentication or API keys for basic usage
• Does NOT log queries or track usage
• Only fetches publicly available price feed data from Pyth Network

Third-Party Services

All price data is provided by Pyth Network. Please refer to Pyth Network's Privacy Policy for information about their data practices.

Local Processing

All data processing occurs locally on your machine. No data is sent to any servers other than the Pyth Network APIs for fetching price information.

---

Support

• Documentation: docs.pyth.network
• Issues: GitHub Issues
• Discord: Pyth Network Discord
• Twitter: @PythNetwork

---

License

Apache-2.0 License - see LICENSE for details.

---

Links

• Pyth Network
• Pyth Documentation
• Hermes API
• Benchmarks API
• MCP Specification