Claude MCPBeginner2 min

Claude MCP + HatiData: Connect in 60 Seconds

Connect Claude Desktop or Claude Code to HatiData via MCP. Get 24 agent-native SQL tools in under a minute.

What You'll Build

A working MCP connection between Claude and HatiData, giving Claude access to 24 tools including SQL queries, memory storage, branch management, and chain-of-thought logging.

Prerequisites

$hati init

$Claude Desktop or Claude Code installed

Architecture

┌──────────────┐    ┌──────────────┐    ┌──────────────┐
│   Claude     │───▶│  HatiData    │───▶│   Engine     │
│  Desktop     │MCP │  MCP Server  │    │  + Vectors   │
└──────────────┘    │  (24 tools)  │    └──────────────┘
                    └──────────────┘

Key Concepts

●MCP (Model Context Protocol): an open standard for connecting AI assistants to external tools and data sources — HatiData implements 24 MCP tools
●Zero-config setup: hati init starts the MCP server automatically alongside the SQL proxy — no separate server to manage
●24 agent-native tools: query, store_memory, search_memory, branch_create, branch_merge, log_reasoning_step, register_trigger, and 17 more
●Works everywhere: the same MCP configuration works with Claude Desktop, Claude Code, Cursor, and any MCP-compatible client

Step-by-Step Implementation

Start HatiData

Initialize and start HatiData locally. The MCP server starts automatically on port 5440.

Bash

# Install and initialize HatiData
curl -fsSL https://hatidata.com/install.sh | sh
hati init

# The MCP server is now running on localhost:5440

Note: hati init starts both the SQL proxy (port 5439) and the MCP server (port 5440).

Configure Claude Desktop

Add HatiData as an MCP server in Claude Desktop's configuration file.

JSON

{
  "mcpServers": {
    "hatidata": {
      "command": "hati",
      "args": ["mcp-server"],
      "env": {
        "HATIDATA_HOST": "localhost",
        "HATIDATA_PORT": "5440"
      }
    }
  }
}

Note: Save this to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or the equivalent path on your OS.

Use HatiData Tools in Claude

Restart Claude Desktop and start using HatiData's 24 MCP tools. Try querying, storing memories, and creating branches.

text

# Example prompts to try in Claude Desktop:

"Query my sales data: SELECT region, SUM(revenue) FROM sales GROUP BY region"

"Store a memory: I prefer dark mode and concise responses"

"Create a branch called 'experiment-1' and run some test queries"

"Show me the chain-of-thought log for today's session"

Expected Output

Claude now has access to 24 HatiData tools: query, store_memory, search_memory, branch_create, log_reasoning_step, and more.

Related Use Case

Operations

Customer Support

Agents That Remember Every Customer

Ready to build?

Install HatiData locally and start building with Claude MCP in minutes.

Join Waitlist