Connecting Claude to Your Business Tools via MCP: A Practical Guide

Connecting Claude to Your Business Tools via MCP: A Practical Guide

Technology Ismaël DIB April 1, 2025 8 min read FR Lire en Français
Claude MCP Business Tools Practical Guide Integration

Practical guide: connecting Claude to your business tools

This guide is for IT teams and Product Owners who want to practically connect Claude to their business tools via MCP, without being development experts. I wrote it after deploying these integrations in my own project context.

Prerequisites

  • Claude Desktop installed (available at claude.ai)
  • Python 3.10+ or Node.js 18+ depending on chosen MCP servers
  • API access to the tools you want to connect

Step 1: Install an existing MCP server

Most popular MCP servers install with a single command. Example with the Jira MCP server:

npm install -g @modelcontextprotocol/server-jira
# or via pip
pip install mcp-server-jira

Step 2: Configure Claude Desktop

In your claude_desktop_config.json file (macOS: ~/Library/Application Support/Claude/, Windows: %APPDATA%/Claude/):

{
  "mcpServers": {
    "jira": {
      "command": "mcp-server-jira",
      "env": {
        "JIRA_URL": "https://your-instance.atlassian.net",
        "JIRA_EMAIL": "you@company.com",
        "JIRA_TOKEN": "your_api_token"
      }
    },
    "slack": {
      "command": "mcp-server-slack",
      "env": {"SLACK_BOT_TOKEN": "xoxb-..."}
    }
  }
}

Step 3: Test the integration

Relaunch Claude Desktop. You'll see a "tools" icon in the interface indicating MCP servers are active. Test with a simple request: "List my open Jira tickets in the EQUANS project".

Recommended MCP servers for project teams

ToolMCP ServerKey capabilities
Jiramcp-server-jiraRead/create/update tickets
Confluencemcp-server-confluenceRead/create pages, search
Slackmcp-server-slackRead/send messages
Google Drivemcp-server-gdriveRead/create files
PostgreSQLmcp-server-postgresNatural language SQL queries
GitHubmcp-server-githubIssues, PRs, code

Security best practices

Never store API tokens in plain text in the config file. Use environment variables or a secrets manager. Apply the principle of least privilege: create tokens with the minimum necessary permissions. Enable MCP logs to audit access.

Recommended MCP servers evaluation
Project team MCP server deployment

Working on an AI automation or digital transformation project?

Let's discuss your challenges. I support IT teams in Switzerland through their AI transition.

Get in touch →