🔧 MCP Tools Setup Guide for Calmhive

Calmhive leverages 70+ powerful MCP (Model Context Protocol) tools to provide advanced AI capabilities. This guide will help you set up the essential tools.

📋 Prerequisites

• Claude Code CLI installed and configured
• Claude Desktop installed (for easy MCP configuration)
• API keys for specific services (see below)

💡 Note: MCP tools integrate with Claude Code CLI. Some tools require their own API keys (like search engines).

🚀 Quick Setup

The easiest way to set up MCP tools is through Claude Desktop:

1. Install Claude Desktop and configure your MCP tools there
2. Import to Claude CLI:

   claude mcp add-from-claude-desktop -s user
   

3. Verify setup:

   calmhive chat "test sequential thinking"
   

🔑 Essential MCP Tools

1. Sequential Thinking Tools (HIGHLY RECOMMENDED)

Advanced problem-solving through structured reasoning.

GitHub: https://github.com/spences10/mcp-sequentialthinking-tools

Setup:

{
  "sequentialthinking": {
    "command": "npx",
    "args": ["-y", "mcp-sequentialthinking-tools"]
  }
}

2. OmniSearch (RECOMMENDED)

Multi-engine search capabilities for research and information gathering.

GitHub: https://github.com/spences10/mcp-omnisearch

Setup (requires API keys):

{
  "omnisearch": {
    "command": "npx",
    "args": ["-y", "mcp-omnisearch"],
    "env": {
      "TAVILY_API_KEY": "tvly-YOUR_KEY_HERE",
      "PERPLEXITY_API_KEY": "pplx-YOUR_KEY_HERE"
    }
  }
}

Get API Keys:

• Tavily: https://tavily.com (free tier available)
• Perplexity: https://www.perplexity.ai/settings/api (requires subscription)

3. Playwright (Browser Automation)

Automated browser testing and web scraping capabilities.

GitHub: https://github.com/automatalabs/mcp-server-playwright

Setup:

{
  "playwright": {
    "command": "npx",
    "args": ["-y", "@automatalabs/mcp-server-playwright"]
  }
}

Features:

• Navigate websites, take screenshots
• Fill forms, click elements
• Extract data from web pages
• Test web applications

4. GitHub (Code Integration)

Access to GitHub repositories and code search.

Setup:

{
  "github": {
    "command": "docker",
    "args": [
      "run", "-i", "--rm", 
      "-e", "GITHUB_PERSONAL_ACCESS_TOKEN",
      "ghcr.io/github/github-mcp-server"
    ],
    "env": {
      "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_YOUR_TOKEN_HERE"
    }
  }
}

Get GitHub Token:

1. Go to https://github.com/settings/tokens
2. Generate new token (classic)
3. Select scopes: repo, read:org

📝 Example Configuration File

Save this as ~/.claude/claude_desktop_config.json:

{
  "mcpServers": {
    "sequentialthinking": {
      "command": "npx",
      "args": ["-y", "mcp-sequentialthinking-tools"]
    },
    "omnisearch": {
      "command": "npx",
      "args": ["-y", "mcp-omnisearch"],
      "env": {
        "TAVILY_API_KEY": "tvly-EXAMPLE_KEY_REPLACE_ME",
        "PERPLEXITY_API_KEY": "pplx-EXAMPLE_KEY_REPLACE_ME"
      }
    },
    "github": {
      "command": "docker",
      "args": [
        "run", "-i", "--rm",
        "-e", "GITHUB_PERSONAL_ACCESS_TOKEN",
        "ghcr.io/github/github-mcp-server"
      ],
      "env": {
        "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_EXAMPLE_TOKEN_REPLACE_ME"
      }
    },
    "gitmcp": {
      "command": "npx",
      "args": ["mcp-remote", "https://gitmcp.io/docs"]
    },
    "Context7": {
      "command": "npx",
      "args": ["-y", "@upstash/context7-mcp@latest"]
    },
    "playwright": {
      "command": "npx",
      "args": ["-y", "@automatalabs/mcp-server-playwright"]
    },
    "figma": {
      "command": "npx",
      "args": [
        "-y", "figma-developer-mcp",
        "--figma-api-key=figd_EXAMPLE_KEY_REPLACE_ME",
        "--stdio"
      ]
    },
    "shadcn-ui": {
      "command": "npx",
      "args": ["-y", "shadcn-ui-mcp-server"]
    }
  }
}

⚠️ Important Notes

1. Replace all example keys with your actual API keys
2. Keep your API keys secure - never commit them to version control
3. Some tools require paid services (e.g., Perplexity API, OpenAI)
4. Start with essentials: Sequential Thinking and OmniSearch are the most important

🧪 Testing Your Setup

After configuration, test each tool:

# Test Sequential Thinking
calmhive chat "use sequential thinking to solve: how to implement a rate limiter"

# Test OmniSearch
calmhive chat "search for the latest React 19 features"

# Test Memento (if configured)
calmhive chat "remember that my favorite framework is Next.js"
calmhive chat "what is my favorite framework?"

🆘 Troubleshooting

"Tool not found" errors:

• Ensure Claude Desktop has the tool configured
• Run claude mcp add-from-claude-desktop -s user again

API key errors:

• Double-check your API keys are valid
• Ensure they have proper permissions

Docker errors (GitHub tool):

• Make sure Docker is installed and running
• Alternative: Use GitMCP instead of GitHub MCP

📚 Additional Resources

• MCP Documentation: https://modelcontextprotocol.io/
• Claude Desktop Setup: https://docs.anthropic.com/en/docs/claude-desktop
• Terminal Velocity Guide: https://github.com/joryeugene/ai-dev-tooling/blob/main/blog/02-terminal-velocity.md#must-use-mcp-tools-amplify-your-ai

---

Pro Tip: You can customize which tools are available by editing config/allowed-tools.json in the Calmhive directory, but the default 86 tools (15 core + 71 MCP integrations) should cover most use cases!