Documentation

gomcptest is a proof of concept (POC) demonstrating how to implement a Model Context Protocol (MCP) with a custom-built host to play with agentic systems.

gomcptest Documentation

Welcome to the gomcptest documentation. This project is a proof of concept (POC) demonstrating how to implement a Model Context Protocol (MCP) with a custom-built host to play with agentic systems.

Documentation Structure

Our documentation follows the Divio Documentation Framework, which organizes content into four distinct types: tutorials, how-to guides, reference, and explanation. This approach ensures that different learning needs are addressed with the appropriate content format.

Tutorials: Learning-oriented content

Tutorials are lessons that take you by the hand through a series of steps to complete a project. They focus on learning by doing, and help beginners get started with the system.

TutorialDescription
Getting Started with gomcptestA complete beginner’s guide to setting up the environment, building tools, and running your first agent with the AgentFlow UI. Perfect for first-time users.
Building Your First OpenAI-Compatible ServerStep-by-step instructions for running and configuring the OpenAI-compatible server with the modern AgentFlow web interface.
Using the cliGCP Command Line Interface⚠️ DEPRECATED: Legacy guide for the cliGCP tool. Use AgentFlow UI instead.

How-to Guides: Problem-oriented content

How-to guides are recipes that guide you through the steps involved in addressing key problems and use cases. They are practical and goal-oriented.

How-to GuideDescription
How to Create a Custom MCP ToolPractical steps to create a new custom tool compatible with the Model Context Protocol, including code templates and examples.
How to Configure the OpenAI-Compatible ServerSolutions for configuring and customizing the OpenAI server with AgentFlow UI, including environment variables, Vertex AI tools, and production setup.
How to Query OpenAI Server with Tool EventsProgrammatically interact with the server using curl, Python, or shell commands to execute tools and monitor their execution events in real-time.
How to Configure the cliGCP Command Line Interface⚠️ DEPRECATED: Legacy configuration guide. Use AgentFlow UI configuration instead.

Reference: Information-oriented content

Reference guides are technical descriptions of the machinery and how to operate it. They describe how things work in detail and are accurate and complete.

ReferenceDescription
Tools ReferenceComprehensive reference of all available MCP-compatible tools, their parameters, response formats, and error handling.
OpenAI-Compatible Server ReferenceTechnical documentation of the server’s architecture, AgentFlow UI, API endpoints, configuration options, and Vertex AI integration.
cliGCP Reference⚠️ DEPRECATED: Legacy CLI reference. Use AgentFlow UI instead.

Explanation: Understanding-oriented content

Explanation documents discuss and clarify concepts to broaden the reader’s understanding of topics. They provide context and illuminate ideas.

ExplanationDescription
gomcptest ArchitectureDeep dive into the system architecture, design decisions, and how the various components interact to create a custom MCP host.
Understanding the Model Context Protocol (MCP)Exploration of what MCP is, how it works, design decisions behind it, and how it compares to alternative approaches for LLM tool integration.
AgentFlow: Modern Web InterfaceComprehensive guide to AgentFlow’s features including tool selection, real-time event notifications, mobile optimization, and conversation management.

Project Components

gomcptest consists of several key components that work together:

Host Components

  • OpenAI-compatible server (host/openaiserver): A server that implements the OpenAI API interface and connects to Google’s Vertex AI for model inference. Includes the modern AgentFlow web UI for interactive chat.
  • cliGCP (host/cliGCP): ⚠️ DEPRECATED - Legacy command-line interface. Use the AgentFlow web UI instead.

AgentFlow Web UI

The modern web-based interface is embedded in the openaiserver binary and provides:

  • Mobile-optimized design with Apple touch icon support
  • Real-time streaming responses via Server-Sent Events
  • Professional styling with accessibility features
  • Conversation management with persistent history
  • File upload support including PDFs
  • Embedded architecture for easy deployment via /ui endpoint

Access AgentFlow by running ./bin/openaiserver and visiting http://localhost:8080/ui

Tools

The tools directory contains various MCP-compatible tools:

File System Operations

  • Bash: Executes bash commands in a persistent shell session
  • Edit: Modifies file content by replacing specified text
  • GlobTool: Finds files matching glob patterns
  • GrepTool: Searches file contents using regular expressions
  • LS: Lists files and directories
  • Replace: Completely replaces a file’s contents
  • View: Reads file contents

Image and Diagram Tools

  • imagen: Generates images using Google’s Imagen API
  • imagen_edit: Edits images using natural language instructions
  • plantuml: Generates PlantUML diagram URLs with syntax validation
  • plantuml_check: Validates PlantUML file syntax

Data and Utility Tools

  • duckdbserver: Provides SQL-based data processing with DuckDB
  • dispatch_agent: Launches specialized sub-agents for specific tasks
  • sleep: Pauses execution for testing and demonstrations

Tutorials

Step-by-step guides to get you started with gomcptest

How-To Guides

Practical guides for solving specific problems with gomcptest

Reference

Technical reference documentation for gomcptest components and tools

Explanation

Understanding-oriented content for gomcptest architecture and concepts