Skip to content

Latest commit

 

History

History
307 lines (220 loc) · 9.21 KB

examples.md

File metadata and controls

307 lines (220 loc) · 9.21 KB

Examples

Real-world usage patterns for the BNB Chain AI Toolkit, organized by difficulty.

How to use this page: Each example lists what you need, shows the code or conversation, and explains what happens. Start with Beginner and work up.

Difficulty What It Means
🟢 Beginner No coding needed, or very simple commands
🟡 Intermediate Some setup required, basic code
🔴 Advanced Multiple components, scripting, or on-chain operations

🟢 Beginner Examples

Check a BNB Balance (No Code)

What you need: BNB Chain MCP server connected to Claude Desktop (see Getting Started)

How it works:

You: What's the BNB balance of 0xF977814e90dA44bFA03b6295A0616a897441aceC?

Claude: That address holds 1,234,567.89 BNB (approximately $768 million at current prices). This appears to be a Binance cold wallet.

Behind the scenes:

Your question → Claude → get_balance tool → BSC RPC → Real balance → Claude's answer

What you learned: Claude didn't guess. It called the MCP server, which queried the actual blockchain, and returned a real number.

Get Today's Crypto Prices

What you need: Market Data component

import { CoinGecko } from '@nirholas/crypto-market-data';

const prices = await CoinGecko.getPrices(['bitcoin', 'binancecoin', 'ethereum']);
console.log(prices);
// Expected output: { bitcoin: 95000, binancecoin: 620, ethereum: 3200 }

What you learned: The market data library fetches real-time prices with one function call. No API key needed for basic usage.

Read the Latest Crypto News

What you need: Crypto News API

# Start the news server
cd market-data/crypto-news && bun start

# Fetch headlines
curl http://localhost:3000/api/news/latest | jq '.[0:5] | .[].title'

Expected output: A list of 5 recent crypto headlines.

Check Market Sentiment (No Code)

What you need: BNB Chain MCP server connected to Claude

You: What's the current crypto market sentiment? Is it a good time to buy?

Claude: The Fear & Greed Index is currently at 35 (Fear). The BNB price is down 3.2% over the past 7 days. Major BNB Chain protocols show stable TVL. While "fear" periods have historically been favorable for long-term buying, this is not financial advice — always assess your own risk tolerance.

What you learned: Claude combines multiple data sources (sentiment index, price history, TVL) into a single coherent answer.

🟡 Intermediate Examples

Set Up a Portfolio Tracking Agent

What you need: Agents + Market Data components Time: 5 minutes

  1. Load the portfolio agent:
cat agents/defi-agents/src/portfolio-analyst.json | jq '.systemRole'

  1. Use it as Claude's system prompt in a Project

  2. Ask questions like:

    • "My portfolio: 10 BNB, 5000 USDT, 2 ETH. How am I doing?"
    • "What's the total value and 24h change?"
    • "Should I rebalance?"

Swap Tokens on PancakeSwap via AI

What you need: BNB Chain MCP + PancakeSwap Trader agent Time: 10 minutes (first time setup), 30 seconds after that

  1. Start the MCP server
  2. Load the PancakeSwap agent
  3. Ask: "Swap 0.1 BNB for CAKE on PancakeSwap"

What happens step by step:

The AI will:

  1. Get the current price quote from PancakeSwap
  2. Calculate expected output and check slippage
  3. Show you the trade details and ask for your confirmation
  4. Only after you say "yes" — execute the swap
  5. Return the transaction hash so you can verify on BscScan

Safety note: The AI always asks for confirmation before executing on-chain transactions. If you're nervous, use testnet first.

Monitor BNB Chain DeFi TVL

What you need: Market Data component Time: 2 minutes

import { DeFiLlama } from '@nirholas/crypto-market-data';

// Get TVL for major BNB Chain protocols
const protocols = ['pancakeswap', 'venus', 'alpaca-finance', 'biswap'];
for (const p of protocols) {
  const tvl = await DeFiLlama.getProtocolTvl(p);
  console.log(`${p}: $${(tvl / 1e9).toFixed(2)}B`);

Expected output:

pancakeswap: $2.15B
venus: $1.83B
alpaca-finance: $0.32B
biswap: $0.08B

}


---

## 🔴 Advanced Examples

### Multi-Server AI Trading Setup

**What you need:** Multiple API keys, Claude Desktop
**Time:** 15-20 minutes

Run all servers together for a complete AI trading terminal:

```json
{
  "mcpServers": {
    "bnbchain": {
      "command": "npx",
      "args": ["-y", "@nirholas/bnbchain-mcp"],
      "env": {
        "BSC_RPC_URL": "https://bsc-dataseed.binance.org",
        "PRIVATE_KEY": "${BSC_PRIVATE_KEY}"
      }
    },
    "binance": {
      "command": "npx",
      "args": ["-y", "@nirholas/binance-mcp"],
      "env": {
        "BINANCE_API_KEY": "${BINANCE_KEY}",
        "BINANCE_SECRET_KEY": "${BINANCE_SECRET}"
      }
    },
    "market": {
      "command": "node",
      "args": ["market-data/crypto-market-data/dist/mcp-server.js"]
    },
    "news": {
      "command": "node",
      "args": ["market-data/crypto-news/dist/mcp-server.js"]
    }
  }
}

Now ask Claude:

  • "What's the sentiment around BNB today? Check the news and fear & greed index."
  • "Show me the best yield opportunities on BSC right now."
  • "If BNB drops below $600, set up a buy order on Binance for 10 BNB."

Convert Any Contract to an AI Tool

Using UCAI to make any BSC contract AI-accessible:

# 1. Get the contract ABI (from BSCScan or your deployment)
curl "https://api.bscscan.com/api?module=contract&action=getabi&address=0xContractAddress" \
  | jq -r '.result' > MyContract.json

# 2. Generate MCP server
cd mcp-servers/ucai
ucai generate --abi ./MyContract.json --chain bsc --output ./my-contract-mcp

# 3. Start it
cd my-contract-mcp && python server.py

# 4. Add to Claude Desktop config

Automated Dust Sweeping Pipeline

#!/bin/bash
# sweep-all-chains.sh — Run weekly via cron

CHAINS=("bsc" "ethereum" "polygon" "arbitrum" "base")
WALLET="0xYourAddress"

for chain in "${CHAINS[@]}"; do
  echo "Sweeping $chain..."
  cd defi-tools/sweep
  bun run sweep --wallet "$WALLET" --chain "$chain" --target USDC --threshold 5
done

echo "Done! All dust swept to USDC."

Register an Agent On-Chain (ERC-8004)

import { ethers } from 'ethers';
import { IdentityRegistry__factory } from '@erc-8004/contracts';

const provider = new ethers.JsonRpcProvider('https://bsc-dataseed.binance.org');
const wallet = new ethers.Wallet(process.env.PRIVATE_KEY, provider);
const registry = IdentityRegistry__factory.connect(REGISTRY_ADDRESS, wallet);

// Register your agent
const tx = await registry.registerAgent(
  'My Trading Agent',
  'Automated BNB/USDT grid trading bot',
  'https://my-agent.example.com/mcp',
  ethers.toUtf8Bytes(JSON.stringify({ version: '1.0', capabilities: ['trade', 'analyze'] }))
);
await tx.wait();
console.log('Agent registered on-chain!');

Integration Patterns

These three patterns represent increasing levels of automation and risk. Start with Pattern 1 and work up as you gain confidence.

Pattern 1: Read-Only Intelligence (Safest)

News API → AI Agent → Human-readable insights

Use when: You want analysis and insights without any risk. No blockchain writes, no trades, no wallet connections.

Example prompts:

  • "Analyze the top 5 BNB Chain protocols by TVL"
  • "What's the market sentiment today?"
  • "Compare PancakeSwap vs Uniswap volume over the past week"

Pattern 2: Monitored Trading (Recommended)

Market Data → AI Agent → Trade Recommendation → Human Approval → Exchange API

Use when: You want AI to find opportunities but you make the final call. AI suggests, human approves. Good for learning and trust-building.

Example prompts:

  • "If BNB drops below $600, draft a limit buy order for me to review"
  • "Find the best yield opportunities on BSC and show me before I deposit"
  • "Scan my portfolio and suggest rebalancing moves — don't execute anything"

Pattern 3: Autonomous Agent (Expert Only)

Market Data + Chain Data → AI Agent → Automated Execution → Exchange/Chain

Use when: You're experienced, understand the risks, and have proper safeguards (slippage limits, position size limits, stop losses). Fully automated — the AI acts without asking.

Example prompts:

  • "Auto-compound my Venus lending rewards every 24 hours"
  • "Run a grid trading strategy: buy every 2% dip, sell every 3% rise, max position 5 BNB"
  • "Sweep dust tokens across all chains weekly and consolidate into USDC on BSC"

Warning: Pattern 3 involves real money moving automatically. Never use Pattern 3 without thoroughly testing on testnet first. Start with tiny amounts. Set strict limits. Monitor logs.

See Also