mcpbeat

Pyth

sendaifun/pyth

Complete guide for Pyth Network - decentralized oracle providing real-time price feeds for DeFi. Covers price feed integration, confidence intervals, EMA prices, on-chain CPI, off-chain fetching, and streaming updates for Solana applications.

30k tokens
context cost
the whole folder, loaded on every use
11
files
ships runnable scripts
0
copies elsewhere
how many repositories repackaged it
122
stars on the repo
on the repository, not the skill itself

Install

one command, takes just this skill from the repository
npx skills add https://github.com/sendaifun/skills --skill pyth

What comes with it

103 440 bytes besides the instruction
docs/troubleshooting.md
examples/on-chain/anchor-integration.rs
examples/on-chain/price-validation.rs
examples/price-feeds/fetch-price.ts
examples/price-feeds/multiple-prices.ts
examples/streaming/real-time-updates.ts
resources/api-reference.md
resources/program-addresses.md
templates/anchor-oracle.rs
templates/pyth-client.ts

The instruction itself

38 sections, as written by the author

Pyth Network Development Guide

Pyth Network is a decentralized oracle providing real-time price feeds for cryptocurrencies, equities, forex, and commodities. This guide covers integrating Pyth price feeds into Solana applications.

Overview

Pyth Network provides:

  • Real-Time Price Feeds - 400ms update frequency with pull oracle model
  • Confidence Intervals - Statistical uncertainty bounds for each price
  • EMA Prices - Exponential moving average prices (~1 hour window)
  • Multi-Asset Support - Crypto, equities, FX, commodities, indices
  • On-Chain Integration - CPI for Solana programs
  • Off-Chain Integration - HTTP and WebSocket APIs via Hermes

Program IDs

| Program | Address | Description |

|---------|---------|-------------|

| Solana Receiver | rec5EKMGg6MxZYaMdyBfgwp4d5rB9T1VQH5pJv5LtFJ | Posts price updates to Solana |

| Price Feed | pythWSnswVUd12oZpeFP8e9CVaEqJg25g1Vtc2biRsT | Stores price feed data |

Deployed on: Solana Mainnet, Devnet, Eclipse Mainnet/Testnet, Sonic networks

| Asset | Hex Feed ID |

|-------|-------------|

| BTC/USD | 0xe62df6c8b4a85fe1a67db44dc12de5db330f7ac66b72dc658afedf0f4a415b43 |

| ETH/USD | 0xff61491a931112ddf1bd8147cd1b641375f79f5825126d665480874634fd0ace |

| SOL/USD | 0xef0d8b6fda2ceba41da15d4095d1da392a0d2f8ed0c6c7bc0f4cfac8c280b56d |

| USDC/USD | 0xeaa020c61cc479712813461ce153894a96a6c00b21ed0cfc2798d1f9a9e9c94a |

| USDT/USD | 0x2b89b9dc8fdf9f34709a5b106b472f0f39bb6ca9ce04b0fd7f2e971688e2e53b |

Full list: https://pyth.network/developers/price-feed-ids

Quick Start

Installation

# TypeScript/JavaScript
npm install @pythnetwork/hermes-client @pythnetwork/pyth-solana-receiver

# Rust (add to Cargo.toml)
# pyth-solana-receiver-sdk = "0.3.0"

Fetch Price (Off-Chain)

import { HermesClient } from "@pythnetwork/hermes-client";

const client = new HermesClient("https://hermes.pyth.network");

const priceIds = [
  "0xe62df6c8b4a85fe1a67db44dc12de5db330f7ac66b72dc658afedf0f4a415b43", // BTC/USD
];

const priceUpdates = await client.getLatestPriceUpdates(priceIds);

for (const update of priceUpdates.parsed) {
  const price = update.price;
  const displayPrice = Number(price.price) * Math.pow(10, price.expo);
  console.log(`Price: $${displayPrice.toFixed(2)}`);
  console.log(`Confidence: ±${Number(price.conf) * Math.pow(10, price.expo)}`);
}

Use Price On-Chain (Rust/Anchor)

use anchor_lang::prelude::*;
use pyth_solana_receiver_sdk::price_update::PriceUpdateV2;

#[derive(Accounts)]
pub struct UsePrice<'info> {
    pub price_update: Account<'info, PriceUpdateV2>,
}

pub fn use_price(ctx: Context<UsePrice>) -> Result<()> {
    let price_update = &ctx.accounts.price_update;
    let clock = Clock::get()?;

    // Get price no older than 60 seconds
    let price = price_update.get_price_no_older_than(
        &clock,
        60, // max age in seconds
    )?;

    msg!("Price: {} × 10^{}", price.price, price.exponent);
    msg!("Confidence: ±{}", price.conf);

    Ok(())
}

Core Concepts

Price Structure

Each Pyth price contains:

| Field | Type | Description |

|-------|------|-------------|

| price | i64 | Price value in fixed-point format |

| conf | u64 | Confidence interval (standard deviation) |

| expo | i32 | Exponent for scaling (e.g., -8 means divide by 10^8) |

| publish_time | i64 | Unix timestamp of price |

Converting to display price:

const displayPrice = price * Math.pow(10, expo);
// Example: price=19405100, expo=-2 → $194,051.00

Confidence Intervals

Confidence intervals represent the uncertainty in the reported price:

// Price is $50,000 ± $50 means:
// - 68% chance true price is between $49,950 - $50,050
// - Use confidence for risk management

const price = 50000;
const confidence = 50;

// Safe lower bound (conservative)
const safeLowerBound = price - confidence;

// Safe upper bound (conservative)
const safeUpperBound = price + confidence;

Best Practice: Reject prices with confidence > 2% of price:

const maxConfidenceRatio = 0.02; // 2%
const confidenceRatio = confidence / Math.abs(price);

if (confidenceRatio > maxConfidenceRatio) {
  throw new Error("Price confidence too wide");
}

EMA Prices

Exponential Moving Average prices smooth out short-term volatility:

  • ~1 hour averaging window (5921 Solana slots)
  • Weighted by inverse confidence (tight confidence = more weight)
  • Good for: liquidations, collateral valuation
  • Available as ema_price and ema_conf
// Use EMA for less volatile applications
const emaPrice = priceUpdate.emaPrice;
const emaConf = priceUpdate.emaConf;

Off-Chain Integration

Hermes Client

Hermes is the recommended way to fetch Pyth prices off-chain.

Public Endpoint: https://hermes.pyth.network

> For production, get a dedicated endpoint from a Pyth data provider.

Fetching Latest Prices

import { HermesClient } from "@pythnetwork/hermes-client";

const client = new HermesClient("https://hermes.pyth.network");

// Single price
const btcPrice = await client.getLatestPriceUpdates([
  "0xe62df6c8b4a85fe1a67db44dc12de5db330f7ac66b72dc658afedf0f4a415b43"
]);

// Multiple prices in one request
const prices = await client.getLatestPriceUpdates([
  "0xe62df6c8b4a85fe1a67db44dc12de5db330f7ac66b72dc658afedf0f4a415b43", // BTC
  "0xff61491a931112ddf1bd8147cd1b641375f79f5825126d665480874634fd0ace", // ETH
  "0xef0d8b6fda2ceba41da15d4095d1da392a0d2f8ed0c6c7bc0f4cfac8c280b56d", // SOL
]);

Streaming Real-Time Updates

import { HermesClient } from "@pythnetwork/hermes-client";

const client = new HermesClient("https://hermes.pyth.network");

const priceIds = [
  "0xe62df6c8b4a85fe1a67db44dc12de5db330f7ac66b72dc658afedf0f4a415b43"
];

// Subscribe to real-time updates via SSE
const eventSource = await client.getPriceUpdatesStream(priceIds, {
  parsed: true,
});

eventSource.onmessage = (event) => {
  const data = JSON.parse(event.data);
  console.log("Price update:", data);
};

eventSource.onerror = (error) => {
  console.error("Stream error:", error);
  eventSource.close();
};

// Close when done
// eventSource.close();

Posting Prices to Solana

import { PythSolanaReceiver } from "@pythnetwork/pyth-solana-receiver";
import { HermesClient } from "@pythnetwork/hermes-client";
import { Connection, Keypair } from "@solana/web3.js";

const connection = new Connection("https://api.mainnet-beta.solana.com");
const wallet = Keypair.fromSecretKey(/* your key */);

const hermesClient = new HermesClient("https://hermes.pyth.network");
const pythReceiver = new PythSolanaReceiver({ connection, wallet });

// Fetch price update data
const priceUpdateData = await hermesClient.getLatestPriceUpdates([
  "0xe62df6c8b4a85fe1a67db44dc12de5db330f7ac66b72dc658afedf0f4a415b43"
]);

// Build transaction to post price
const transactionBuilder = pythReceiver.newTransactionBuilder();
await transactionBuilder.addPostPriceUpdates(priceUpdateData.binary.data);

// Add your program instruction that uses the price
// transactionBuilder.addInstruction(yourInstruction);

// Send transaction
const transactions = await transactionBuilder.buildVersionedTransactions({
  computeUnitPriceMicroLamports: 50000,
});

for (const tx of transactions) {
  const sig = await connection.sendTransaction(tx);
  console.log("Transaction:", sig);
}

On-Chain Integration (Rust)

Setup

Add to Cargo.toml:

[dependencies]
pyth-solana-receiver-sdk = "0.3.0"
anchor-lang = "0.30.1"

Reading Price in Anchor Program

use anchor_lang::prelude::*;
use pyth_solana_receiver_sdk::price_update::{PriceUpdateV2, get_feed_id_from_hex};

declare_id!("YourProgramId...");

// BTC/USD price feed ID
const BTC_USD_FEED_ID: &str = "0xe62df6c8b4a85fe1a67db44dc12de5db330f7ac66b72dc658afedf0f4a415b43";

#[program]
pub mod my_program {
    use super::*;

    pub fn check_price(ctx: Context<CheckPrice>) -> Result<()> {
        let price_update = &ctx.accounts.price_update;
        let clock = Clock::get()?;

        // Verify this is the correct feed
        let feed_id = get_feed_id_from_hex(BTC_USD_FEED_ID)?;

        // Get price no older than 60 seconds
        let price = price_update.get_price_no_older_than_with_custom_verification(
            &clock,
            60,
            &feed_id,
            ctx.accounts.price_update.to_account_info().owner,
        )?;

        msg!("BTC/USD Price: {} × 10^{}", price.price, price.exponent);
        msg!("Confidence: ±{}", price.conf);

        Ok(())
    }
}

#[derive(Accounts)]
pub struct CheckPrice<'info> {
    #[account(
        constraint = price_update.to_account_info().owner == &pyth_solana_receiver_sdk::ID
    )]
    pub price_update: Account<'info, PriceUpdateV2>,
}

Using Price for Calculations

pub fn swap_with_oracle(
    ctx: Context<SwapWithOracle>,
    amount_in: u64,
) -> Result<()> {
    let price_update = &ctx.accounts.price_update;
    let clock = Clock::get()?;

    // Get price with staleness check
    let price = price_update.get_price_no_older_than(&clock, 30)?;

    // Validate confidence (max 1% of price)
    let conf_ratio = (price.conf as u128 * 10000) / (price.price.unsigned_abs() as u128);
    require!(conf_ratio <= 100, ErrorCode::ConfidenceTooWide);

    // Convert price to usable format
    // price.price is in fixed-point with price.exponent
    let price_scaled = if price.exponent >= 0 {
        (price.price as u128) * 10_u128.pow(price.exponent as u32)
    } else {
        (price.price as u128) / 10_u128.pow((-price.exponent) as u32)
    };

    // Calculate output amount using oracle price
    let amount_out = (amount_in as u128)
        .checked_mul(price_scaled)
        .ok_or(ErrorCode::MathOverflow)?
        / 1_000_000; // Adjust for decimals

    msg!("Swap {} -> {} using price {}", amount_in, amount_out, price_scaled);

    Ok(())
}

#[error_code]
pub enum ErrorCode {
    #[msg("Price confidence interval too wide")]
    ConfidenceTooWide,
    #[msg("Math overflow")]
    MathOverflow,
}

Multiple Price Feeds

#[derive(Accounts)]
pub struct Liquidation<'info> {
    #[account(
        constraint = collateral_price.to_account_info().owner == &pyth_solana_receiver_sdk::ID
    )]
    pub collateral_price: Account<'info, PriceUpdateV2>,

    #[account(
        constraint = debt_price.to_account_info().owner == &pyth_solana_receiver_sdk::ID
    )]
    pub debt_price: Account<'info, PriceUpdateV2>,
}

pub fn check_liquidation(ctx: Context<Liquidation>) -> Result<bool> {
    let clock = Clock::get()?;

    let collateral = ctx.accounts.collateral_price
        .get_price_no_older_than(&clock, 60)?;
    let debt = ctx.accounts.debt_price
        .get_price_no_older_than(&clock, 60)?;

    // Normalize to same exponent for comparison
    let collateral_value = normalize_price(collateral.price, collateral.exponent);
    let debt_value = normalize_price(debt.price, debt.exponent);

    // Check if undercollateralized
    let is_liquidatable = collateral_value < debt_value * 150 / 100; // 150% ratio

    Ok(is_liquidatable)
}

fn normalize_price(price: i64, expo: i32) -> i128 {
    let target_expo = -8; // Normalize to 8 decimals
    let adjustment = expo - target_expo;

    if adjustment >= 0 {
        (price as i128) * 10_i128.pow(adjustment as u32)
    } else {
        (price as i128) / 10_i128.pow((-adjustment) as u32)
    }
}

Best Practices

1. Always Check Staleness

// Don't use old prices - set appropriate max age
let max_age_seconds = 60;
let price = price_update.get_price_no_older_than(&clock, max_age_seconds)?;

2. Validate Confidence Intervals

// Reject prices with wide confidence (high uncertainty)
const MAX_CONF_BPS: u64 = 200; // 2%

let conf_bps = (price.conf as u128 * 10000) / (price.price.unsigned_abs() as u128);
require!(conf_bps <= MAX_CONF_BPS as u128, ErrorCode::ConfidenceTooWide);

3. Verify Account Ownership

// Always verify the price account is owned by Pyth
#[account(
    constraint = price_update.to_account_info().owner == &pyth_solana_receiver_sdk::ID
)]
pub price_update: Account<'info, PriceUpdateV2>,

4. Use EMA for Sensitive Operations

// For liquidations, use EMA to avoid manipulation
let ema_price = price_update.get_ema_price_no_older_than(&clock, 60)?;

5. Handle Price Unavailability

try {
  const price = await client.getLatestPriceUpdates([feedId]);
  // Use price
} catch (error) {
  // Fallback behavior or reject transaction
  console.error("Price unavailable:", error);
}

6. Consider Frontrunning

  • Adversaries may see price updates before your transaction
  • Don't design logic that races against price updates
  • Use appropriate slippage tolerances

Price Feed Types

Fixed Price Feed Accounts

  • Maintained continuously by Pyth
  • Fixed address per feed
  • Always has most recent price
  • Shared by all users (potential congestion)

Ephemeral Price Update Accounts

  • Created per transaction
  • Can specify shard ID for parallelization
  • Rent can be recovered after use
  • Better for high-throughput applications
// Use shard ID to avoid congestion
const transactionBuilder = pythReceiver.newTransactionBuilder({
  shardId: Math.floor(Math.random() * 65536), // Random shard
});

Resources

Official Documentation

GitHub Repositories

NPM Packages

Rust Crates


Skill Structure

pyth/
├── SKILL.md                          # This file
├── resources/
│   ├── program-addresses.md          # All program IDs and feed IDs
│   └── api-reference.md              # SDK API reference
├── examples/
│   ├── price-feeds/
│   │   ├── fetch-price.ts            # Basic price fetching
│   │   └── multiple-prices.ts        # Multiple price feeds
│   ├── on-chain/
│   │   ├── anchor-integration.rs     # Anchor program example
│   │   └── price-validation.rs       # Price validation patterns
│   └── streaming/
│       └── real-time-updates.ts      # WebSocket streaming
├── templates/
│   ├── pyth-client.ts                # TypeScript client template
│   └── anchor-oracle.rs              # Anchor program template
└── docs/
    └── troubleshooting.md            # Common issues and solutions

How to use it

Copy the folder

Take sendaifun/pyth from the repository into ~/.claude/skills for personal use, or into .claude/skills inside a project.

Check the name does not clash

The agent identifies a skill by the name field in its header. Two skills with the same name cannot sit side by side — one of them will be ignored.

Install what it needs

The instructions reference npm. Without those the skill loads but fails at the first command.