> ## Documentation Index
> Fetch the complete documentation index at: https://docs.chainstream.io/llms.txt
> Use this file to discover all available pages before exploring further.
# AI Agent with MCP
> Build an AI trading assistant with Claude that can query on-chain data
This tutorial introduces how to use ChainStream MCP Server to build an AI assistant that can query on-chain data, analyze markets, and assist with trading decisions.
**Estimated Time**: 30 minutes\
**Difficulty Level**: ⭐⭐ Beginner
***
## Objective
Build an AI assistant with on-chain awareness:
```mermaid theme={null}
flowchart TD
A[User Query] --> B[Claude + MCP]
B --> C[ChainStream API]
C --> B
B --> D[Response]
```
**Capabilities**:
* ✅ Query Token prices and balances
* ✅ Analyze wallet holdings
* ✅ Track Smart Money
***
## Step 1: Configure MCP
### Prerequisites
* ChainStream account (to get Client ID and Client Secret)
* Claude Desktop or Cursor
* Node.js 18+
### Install MCP Server
```bash theme={null}
npm install -g @chainstream/mcp-server
```
### Configure Claude Desktop
Edit the configuration file:
```bash theme={null}
# Config file path
~/Library/Application Support/Claude/claude_desktop_config.json
```
```bash theme={null}
# Config file path
%APPDATA%\Claude\claude_desktop_config.json
```
Add configuration:
```json theme={null}
{
"mcpServers": {
"chainstream": {
"url": "https://mcp.chainstream.io/mcp",
"headers": {
"Authorization": "Bearer your_access_token"
}
}
}
}
```
Restart Claude Desktop.
For detailed steps, refer to the [MCP Setup Guide](/en/docs/ai-agents/mcp-server/setup).
***
## Step 2: Conversation Examples
### Example 1: Check ETH Price
```
User: Show me the ETH price
Claude: [Calls get_token_price tool]
ETH Current Price:
- Price: $2,015.32
- 24h Change: +2.5%
- 24h Volume: $12.5B
- Market Cap: $242B
```
### Example 2: Analyze Wallet Holdings
```
User: Analyze the holdings of 0xd8dA6BF26964aF9D7eEd9e03E53415D37aA96045
Claude: [Calls get_token_balance and analyze_wallet tools]
This address (vitalik.eth) holdings analysis:
Holdings Overview
| Token | Amount | Value |
|-------|--------|-------|
| ETH | 1,234 | $2.48M |
| USDC | 50,000 | $50K |
| ENS | 10,000 | $25K |
Wallet Characteristics
- Wallet Age: 7+ years
- Tags: Core Developer, Long-term Holder
- Main Activities: DeFi, NFT, Donations
```
### Example 3: Track Smart Money
```
User: Which Smart Money addresses have been buying ARB recently?
Claude: [Calls get_smart_money_activities tool]
Past 24h Smart Money ARB Buy Records:
1. 0xabc...123 (Smart Score: 92)
- Bought: 500,000 ARB ($450K)
- Time: 2 hours ago
- Tags: DeFi Expert, High Win-rate Trader
2. 0xdef...456 (Smart Score: 85)
- Bought: 200,000 ARB ($180K)
- Time: 5 hours ago
- Tags: Institutional Wallet
Trend Analysis: Smart Money overall shows net buying stance on ARB
```
***
## Step 3: Advanced Usage
### Multi-step Analysis
```
User: First check this address's holdings, then analyze its recent trades,
and determine its investment style
Claude: I'll analyze step by step:
Step 1: Holdings Query
[Calls get_token_balance]
...
Step 2: Transaction History
[Calls get_transaction_history]
...
Step 3: Comprehensive Analysis
[Calls analyze_wallet]
Based on the above data, this address's investment style is:
- Type: Medium to Long-term Holder
- Preference: DeFi Blue Chips + Layer2
- Risk Appetite: Moderate
- Characteristic: Good at adding positions during pullbacks
```
### Comparative Analysis
```
User: Compare the trading strategies of these two addresses
Claude: [Calls analyze_wallet x2]
| Dimension | Address A | Address B |
|-----------|-----------|-----------|
| Trading Frequency | High (50/day) | Low (5/day) |
| Holding Style | Short-term | Long-term |
| Win Rate | 62% | 78% |
| Preferred Sector | Meme Coins | DeFi |
| Risk Level | High | Medium |
```
***
## Security Considerations
### Strategy Limitations
The AI assistant only provides data queries and analysis, not direct trade execution.
```
User: Buy 100 USDC worth of ETH for me
Claude: I can help you query the current ETH price and best trading routes,
but cannot execute trades directly.
Current ETH price: $2,015
Suggested execution methods:
1. Uniswap V3 - Expected slippage 0.1%
2. 1inch Aggregator - May get better price
Please execute the trade manually in your wallet.
```
### Human Confirmation
For important decisions, AI provides analysis but recommends human confirmation:
Based on analysis, multiple Smart Money addresses are buying ARB, but please note:
* This does not constitute investment advice
* Markets are risky, please judge independently
* Recommend combining with more information sources
***
## Extension: Custom Agent
### Build with MCP SDK
```javascript theme={null}
import Anthropic from '@anthropic-ai/sdk';
import { Client } from '@modelcontextprotocol/sdk/client/index.js';
import { SSEClientTransport } from '@modelcontextprotocol/sdk/client/sse.js';
const anthropic = new Anthropic();
// Initialize MCP Client
async function createMcpClient() {
const transport = new SSEClientTransport(
new URL('https://mcp.chainstream.io/sse'),
{
requestInit: {
headers: {
'Authorization': `Bearer ${process.env.CHAINSTREAM_ACCESS_TOKEN}`
}
}
}
);
const client = new Client({
name: 'custom-agent',
version: '1.0.0'
});
await client.connect(transport);
return client;
}
async function chat(userMessage) {
const mcp = await createMcpClient();
// Get available tools
const { tools } = await mcp.listTools();
// Convert to Anthropic format
const anthropicTools = tools.map(tool => ({
name: tool.name,
description: tool.description,
input_schema: tool.inputSchema
}));
// Call Claude
let response = await anthropic.messages.create({
model: 'claude-sonnet-4-20250514',
max_tokens: 1024,
tools: anthropicTools,
messages: [{ role: 'user', content: userMessage }]
});
// Handle tool calls
while (response.stop_reason === 'tool_use') {
const toolUse = response.content.find(c => c.type === 'tool_use');
// Call MCP tool
const toolResult = await mcp.callTool({
name: toolUse.name,
arguments: toolUse.input
});
// Continue conversation with tool result
response = await anthropic.messages.create({
model: 'claude-sonnet-4-20250514',
max_tokens: 1024,
tools: anthropicTools,
messages: [
{ role: 'user', content: userMessage },
{ role: 'assistant', content: response.content },
{
role: 'user',
content: [{
type: 'tool_result',
tool_use_id: toolUse.id,
content: JSON.stringify(toolResult.content)
}]
}
]
});
}
await mcp.close();
return response.content.find(c => c.type === 'text')?.text;
}
// Usage
const result = await chat('Check ETH price on Ethereum');
console.log(result);
```
***
## Available MCP Tools
| Tool | Description |
| ---------------------------- | --------------------------------- |
| `get_token_balance` | Query address token balance |
| `get_token_price` | Get real-time token price |
| `get_transaction_history` | Query address transaction history |
| `get_smart_money_activities` | Get Smart Money activities |
| `get_defi_positions` | Query DeFi positions |
| `analyze_wallet` | Analyze wallet characteristics |
| `get_token_holders` | Get token holder analysis |
***
## FAQ
1. Confirm MCP Server is configured correctly
2. Restart Claude Desktop
3. Check if Client ID and Client Secret are valid
4. Try more explicit prompts like "Use ChainStream to query..."
On-chain data queries take time, especially for complex analysis. You can:
1. Ask step by step to reduce single query load
2. Specify a specific chain to reduce query scope
The current version only supports data queries, not direct trade execution. This is for security reasons. For automated trading, we recommend executing through a separate trading system based on AI analysis results.
***
## Related Documentation
Detailed configuration steps
View all available tools