Atlas-G Protocol

Agentic Portfolio System - A compliance-grade MCP server that serves as both human and machine-readable portfolio.

Python FastAPI Cloud Run MCP

🎯 Overview

Atlas-G Protocol transforms a traditional developer portfolio into an autonomous agent that demonstrates compliance-grade engineering in real-time. Instead of reading about experience with "strict state management" and "hallucination mitigation," users interact with an agent that actively demonstrates these capabilities.

Key Features

MCP Server: Machine-readable portfolio accessible by AI development environments
Governance Layer: Real-time hallucination mitigation via knowledge graph validation
Live Audit Log: Streams internal compliance checks to the UI
WebSocket Streaming: Real-time "Thought-Action" loop visualization
CSP Headers: Configured for DEV.to iframe embedding

🔒 Privacy & Data Governance

The Atlas-G Protocol follows a "Private-by-Design" pattern to ensure sensitive career data isn't leaked in public repositories:

Template Pattern: All proprietary information (work history, PII) is stored in data/resume.txt, which is explicitly excluded from the repository via .gitignore.
resume.template.txt: A sanitized template is provided for open-source users to populate with their own data.
Hallucination Mitigation: The agent's governance layer validates every claim against the local resume.txt knowledge graph before responding.

🏗️ Architecture

┌─────────────────────────────────────────────────────┐
│                   Cloud Run Instance                 │
├─────────────────────────────────────────────────────┤
│  ┌─────────────────┐    ┌─────────────────────────┐ │
│  │  React Frontend │◄──►│  FastAPI Backend        │ │
│  │  (Terminal UI)  │    │  - Agent Core           │ │
│  └─────────────────┘    │  - Governance Layer     │ │
│                         │  - MCP Server           │ │
│                         └───────────┬─────────────┘ │
│                                     │               │
│                         ┌───────────▼─────────────┐ │
│                         │  Tools                  │ │
│                         │  - query_resume         │ │
│                         │  - verify_employment    │ │
│                         │  - audit_project        │ │
│                         └─────────────────────────┘ │
└─────────────────────────────────────────────────────┘

🚀 Quick Start

Prerequisites

Python 3.11+
Google Cloud API Key (for Gemini)

Installation

# Clone the repository
cd Atlas-G\ Protocol

# Create virtual environment
python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

# Install dependencies
pip install -e ".[dev]"

# Copy environment template
cp .env.example .env
# Edit .env with your GOOGLE_API_KEY

Run Locally

# Start the server
uvicorn backend.main:application --reload --port 8080

# Open http://localhost:8080

Run Tests

pytest backend/tests/ -v

🔧 MCP Integration

Connect your AI development environment to the Atlas-G MCP server:

{
  "mcpServers": {
    "atlas-g-protocol": {
      "command": "python",
      "args": ["-m", "backend.mcp_server"]
    }
  }
}

Available Tools

Tool	Description
`query_resume`	Semantic search over resume knowledge graph
`verify_employment`	Cross-reference employment claims
`audit_project`	Deep-dive into project architecture

☁️ Deploy to Cloud Run

gcloud run deploy atlas-g-portfolio \
  --source . \
  --allow-unauthenticated \
  --region us-central1 \
  --labels dev-tutorial=devnewyear2026 \
  --set-env-vars GOOGLE_API_KEY=your_key_here

📁 Project Structure

Atlas-G Protocol/
├── backend/
│   ├── __init__.py
│   ├── main.py          # FastAPI application
│   ├── agent.py         # Thought-Action loop
│   ├── governance.py    # Hallucination mitigation
│   ├── mcp_server.py    # FastMCP wrapper
│   ├── config.py        # Settings management
│   └── tools/
│       ├── resume_rag.py
│       └── verification.py
├── frontend/            # React UI (Phase 3)
├── data/
│   └── resume.txt       # Knowledge graph source
├── Dockerfile
├── pyproject.toml
└── mcp_config.json

🔒 Security

CSP Headers: frame-ancestors 'self' https://dev.to https://*.dev.to
Governance Layer: All AI responses validated against resume data
PII Detection: Automatic filtering of sensitive information
Jailbreak Protection: Pattern-based detection and blocking

📄 License

MIT License - See LICENSE for details.

📢 Credits

Audio: Emergency Alarm.wav by Mozfoo (CC0)

atlas-g-protocol