Back to the Campfire Blog

How to Create Your Own MCP Server

Written by: Liz Bowers Tags: technical-seo, data-pipelines, template

Published: Mar 23, 2026  |  Last Updated: Mar 24, 2026

A basic Model Context Protocol (MCP) server can make tasks more reliable and data-driven. Follow this guide and learn how to build something that works for you and your data.

> To see what I've built using this server, watch my product demo in the Claude Code showcase.

> Check out Noah Learner's AirOps webinar to get the video walkthrough of what it looks like inside the terminal. 

Architecture diagram of a basic MCP server

Key tech you need 

• Runtime: Node.js or Python
• MCP: Claude Desktop/Code with MCP server support
• APIs: Google Cloud (GA4, GSC, Sheets), third-party APIs
• Storage: Google Sheets, GitHub, local filesystem
• Authentication: OAuth 2.0, service accounts, API keys

How to recreate this build on your own

Phase 1: Foundation setup

1. Set up Google Cloud Project

• Create a project in Google Cloud Console
• Enable APIs: Analytics API, Search Console API, Sheets API, and anything else you need to connect
• Create service accounts with proper permissions
• Download service account JSON credentials

2. Configure MCP Servers

• Install Claude Desktop or Claude Code CLI
• Configure MCP settings for Google Analytics 4
• Set up Search Console MCP server (or use direct API)
• Store credentials securely

3. Set up Third-Party APIs

• Login/ Sign-up for third-party tool connections
• Obtain API keys
• Store in local .env file (never commit)

Phase 2: Data integration

4. Create data source connections

• Build a service layer for each data source
• Implement authentication handlers
• Create data fetching functions with error handling
• Add rate limiting and retry logic

5. Set up Google Sheets integration

• Create Google spreadsheets with a smart naming convention (if this is what you're using as a source or output destination)
• Build export functions using Google Sheets API
• Implement sheet creation and formatting logic
• Add data validation and formatting

6. Create local file references

• Set up local files
• Build local file reading capabilities

Phase 3: Build processing engine

7. Create data aggregation layer

• Build unified data model for all sources
• Create aggregation functions
• Implement data normalization and cleaning
• Add date range handling

Phase 4: Build agentic workflows

Connect tasks together that use your live data as triggers for workflows. 

Phase 5: Automation & integration

8. Create Command-line interface (CLI) commands

• Build CLI for each workflow
• Add configuration options
• Implement interactive prompts
• Add logging and error reporting

9. Add version control

• Set up GitHub repository to store versions of your builds to revert to or reference later if needed
• Create branch strategy so you can create variations and replications

Questions? Hit me up on Linkedin or inside the community.