Blockchain wallet analytics MCP

MCP server for retriving on-chain wallet swap transactions. Fetches and parses swap data from blockchain wallets for LLM based analysis.

Features

• DEX detection — Identifies Raydium, Jupiter, Meteora, Orca, Pump.fun from transaction instructions

• Token classification — Auto-labels stablecoins, base currencies, and meme tokens

• Transaction categories — Classifies swaps, transfers, staking, NFT mints

• Multi-hop support — Aggregates complex swap paths into single entries

• Currency-agnostic filters — min_amount_sent/received work on any token, not just SOL

• RPC auto-detection — Single SOLANA_RPC_URL config; detects provider from URL

Related MCP server: onchainos MCP Server

Supported Chains

The parser is chain-agnostic in design — swap detection logic applies to any EVM or Solana wallet. New chains are added by extending the transaction parsing module.

Use Cases

Installation

pip install wallet-analytics-mcp

Or via uv:

uv pip install wallet-analytics-mcp

Quick Start

Run without installing using uvx:

uvx wallet-analytics-mcp

Configuring an MCP Client

Add to your client's MCP configuration. Example for Claude Desktop (~/.claude/claude_desktop_config.json):

{
  "mcpServers": {
    "wallet-analytics": {
      "command": "uvx",
      "args": ["wallet-analytics-mcp"],
      "env": {
        "SOLANA_RPC_URL": "https://your-rpc-endpoint"
      }
    }
  }
}

The env block passes variables directly to the server subprocess — no need to export them globally. Other MCP-compatible clients (Cursor, Windsurf, Cline, etc.) use the same JSON config format.

Configuration

All settings via environment variables:

Point SOLANA_RPC_URL at any provider — Alchemy, Helius, QuickNode, etc. Falls back to Solana public if not set.

Available Tools

get_raw_transactions

Fetch all detected swap transactions for a wallet address. By default returns every swap regardless of token pair. Optional filters let the LLM narrow results.

Parameters:

Returns: List of swaps with token pairs, quantities, timestamps, signatures, DEX platform, and transaction category. Response includes filters_applied showing which filters were active.

Design

The parser returns all detected swaps — no hard-coded drops. Filtering happens at the tool level so the LLM decides what's relevant. Each swap includes:

• Platform — Detected DEX program (Raydium, Jupiter, Meteora, Orca, Pump.fun) from transaction instructions

• Category — Transaction type (swap, transfer, staking, nft, other) based on program IDs called

• Token Symbols — Human-readable symbols for known tokens (e.g., BONK, WIF, POPCAT)

This catches USDC→meme swaps, meme-to-meme trades, and any other token pair.

Attribution

This project was generated by Qwen3.6-27B (Q3_K_M quant) running locally via llama.cpp. Key inference settings:

• Context: 65,536 tokens

• Temperature: 0.6, Top-K: 20, Top-P: 0.95

• Presence penalty: 1.5

• Flash attention enabled, Q8_0 KV cache

License

MIT