Agentic MCP Solution

A practical, extensible backend for building agentic AI systems that connect language models to real tools and workflows.

Why this project matters

This repository is a strong example of how I approach modern AI engineering: combining language models, tool use, and structured backend development to create systems that feel useful beyond demos. It reflects a hands-on interest in building intelligent applications that can interact with real services such as Gmail, Google Docs, and AI providers in a controlled and modular way.

Related MCP server: mcp-google

What I built

This project demonstrates a full-stack mindset for agentic systems with:

• a FastAPI backend for exposing capabilities and orchestrating workflows

• a CLI interface for developer-friendly interaction and testing

• modular MCP server integrations for external tools and services

• support for Google Workspace operations and LLM-backed reasoning

• a configuration-driven architecture that is easy to extend

Engineering highlights

1. Agentic architecture

The system is designed around the idea that agents should not be limited to text generation. Instead, they should be able to discover capabilities, invoke tools, and complete practical tasks through structured interfaces.

2. Modular integration design

Each capability is isolated behind an MCP-style server abstraction, which makes the platform extensible and easier to evolve as new services are added.

3. Production-minded backend structure

The project uses:

• FastAPI and Pydantic for API design and validation

• async patterns for scalable workflow execution

• OAuth-based Google authentication flows

• environment and YAML-based configuration for runtime flexibility

4. Developer experience

A CLI interface is included so that capabilities can be tested and used quickly, which is especially valuable for iterative development and demonstrating agent behavior.

Skills this project showcases

This repository highlights a broad set of skills that are relevant to both recruiters and fellow developers:

• Python backend development

• API design and service orchestration

• LLM tool-use patterns and agentic workflows

• integration with external APIs and cloud services

• modular system architecture

• asynchronous programming concepts

• authentication and secure configuration handling

• practical software design for extensibility

Example workflow

A simple example of the project’s purpose is an agent that can:

1. inspect Gmail messages

2. analyze the contents with an LLM

3. generate a structured report

4. create or update a Google Doc with the result

That is the kind of real-world, tool-using workflow this project is designed to support.

Architecture at a glance

The repository is organized into a few clear layers:

• src/api – FastAPI entrypoints and request handling

• src/cli – command-line interaction layer

• src/core – shared orchestration and authentication logic

• src/mcp_servers – service-specific integrations such as Gmail and Google Docs

• config – runtime configuration and settings

• tests – test coverage for the application surface

Getting started

Prerequisites

• Python 3.13+

• uv (recommended) or pip

• Google OAuth credentials in credentials/credentials.json

• environment variables for Gemini and Anthropic access

Install dependencies

uv sync

Or:

pip install -e .

Configure environment variables

export GEMINI_API_KEY="your_gemini_key"
export ANTHROPIC_API_KEY="your_anthropic_key"
export API_HOST="0.0.0.0"
export API_PORT="8000"
export DEBUG="false"

Start the API

uv run agentic-api

Use the CLI

uv run agentic-cli capabilities
uv run agentic-cli health

Project structure

agentic-mcp-solution/
├── config/                # runtime configuration
├── credentials/           # OAuth and service credentials
├── docs/                  # project documentation and notes
├── scripts/               # utility scripts
├── src/                   # application source code
│   ├── api/               # FastAPI API layer
│   ├── cli/               # CLI interface
│   ├── core/              # orchestration and shared logic
│   └── mcp_servers/       # external integrations
├── tests/                 # automated tests
├── main.py                # package entrypoint
├── pyproject.toml         # project dependencies and metadata
└── README.md              # project documentation

What this project says about my approach

This project reflects a practical engineering style that values:

• clarity over complexity

• modularity over monolithic implementation

• real-world usefulness over theoretical demos

• thoughtful integration over flashy prototypes

It is the kind of work I would be excited to build on further in production-focused AI and backend systems.

License

This project is licensed under the MIT License.