Hermes
An AI-powered developer productivity server that enables managing emails, analyzing GitHub repositories, searching AI/ML papers, and tailoring resumes via MCP tools.
README
<div align="center">
<br/>
<img src="https://img.shields.io/badge/HERMES-1.0-f0a05c?style=for-the-badge&labelColor=0a0a0f&color=f0a05c" alt="Hermes 1.0"/>
<br/><br/>
An intelligent developer productivity platform powered by AI.<br/> Manage email, analyse GitHub repositories, and tailor resumes — all from a single local server.
<br/>
<br/>
</div>
Overview
Hermes is a dual-interface AI productivity server that runs entirely on your local machine. It exposes five AI-powered tools through two interfaces simultaneously — a Model Context Protocol (MCP) server for Claude Desktop, and a Flask HTTP API backing a built-in web UI.
The server is powered by Groq (llama-3.3-70b-versatile) for all AI tasks, with Waitress (Windows) or Gunicorn (Linux/macOS) as the production WSGI layer — auto-selected at runtime with zero configuration.
No data leaves your machine except for API calls to Gmail, GitHub, Groq, and arXiv/HuggingFace/PapersWithCode.
Capabilities
| Module | Description |
|---|---|
| Inbox Sorter | Fetches Gmail and classifies every message by priority — critical, high, medium, or low — using AI analysis of subject, sender, and content. |
| Email Composer | Generates professional emails via Groq with configurable tone. Supports one-click send via Gmail API. |
| AI/ML Search | Deep research across arXiv, HuggingFace, and PapersWithCode. Returns ranked papers, models, and structured insights. |
| GitHub Analyzer | Full portfolio analysis across 8 actions: repo overview, commit activity, README quality scoring, stale repo detection, AI code review, tech stack mapping, and dependency auditing. |
| Resume Tailor | Two-phase AI tailoring — JD analysis followed by resume rewriting with match scoring, gap analysis, and interview tips. |
Architecture

Full data flow from Claude Desktop and Browser UI through the MCP/HTTP layers, tools, services, and external APIs.
Project Structure
hermes/
├── main.py # Flask app + MCP server (WSGI entry point)
├── serve.py # Smart launcher — auto-detects OS and WSGI server
├── gunicorn.conf.py # Gunicorn configuration (Linux/macOS)
├── logger.py # Structured logging
├── check_groq.py # API key diagnostic utility
│
├── tools/
│ ├── mail_fetcher.py # Gmail fetch + Groq classification
│ ├── mail_writer.py # Email generation + Gmail send
│ ├── ai_search.py # Multi-source AI/ML research
│ ├── github_analyzer.py # GitHub analysis — 8 actions
│ └── resume_tailor.py # Two-phase resume tailoring
│
├── services/
│ ├── claude_service.py # Groq client + shared AI helpers
│ ├── gmail_service.py # Gmail OAuth 2.0 + API wrapper
│ └── github_service.py # PyGithub wrapper — 8 analysis functions
│
├── ui/
│ └── index.html # Single-file dark UI — no build step required
│
├── tests/
│ ├── test_imports.py # Module import smoke tests
│ ├── test_flask_routes.py # HTTP endpoint tests (mocked)
│ ├── test_github_analyzer.py # GitHub tool unit tests
│ └── test_resume_tailor.py # Resume tailor unit tests
│
├── .github/
│ └── workflows/
│ └── ci.yml # CI pipeline — lint, imports, unit tests
│
├── .env # Local secrets — never committed
├── .env.example # Environment variable template
└── requirements.txt
Getting Started
Prerequisites
- Python 3.11 or higher
- Groq API key — free tier sufficient
- Gmail OAuth 2.0 credentials (
credentials.jsonfrom Google Cloud Console) - GitHub Personal Access Token with
reposcope
Installation
git clone https://github.com/rayyan666/MAIL-MCP.git hermes
cd hermes
python -m venv venv
venv\Scripts\activate # Windows
# source venv/bin/activate # Linux / macOS
pip install -r requirements.txt
Configuration
cp .env.example .env
Edit .env:
GROQ_API_KEY=gsk_your_key_here
GH_TOKEN=ghp_your_token_here
Place credentials.json (Gmail OAuth) in the project root.
Verify Setup
python check_groq.py
Run
python serve.py
Open http://localhost:5000 in your browser.
Note: Run from a plain terminal window rather than the VS Code integrated terminal to ensure
.envvariables load correctly viapython-dotenv.
Run Modes
| Command | Description |
|---|---|
python serve.py |
Production HTTP server — Waitress on Windows, Gunicorn on Linux |
python serve.py --mcp |
MCP stdio mode for Claude Desktop with HTTP server running in background |
python serve.py --dev |
Flask development server with debug mode enabled |
PORT=8080 python serve.py |
Run on a custom port |
MCP Integration
To use Hermes as an MCP server with Claude Desktop, add the following to %APPDATA%\Claude\claude_desktop_config.json:
{
"mcpServers": {
"hermes": {
"command": "C:\\path\\to\\hermes\\venv\\Scripts\\python.exe",
"args": ["C:\\path\\to\\hermes\\serve.py", "--mcp"],
"env": {
"GROQ_API_KEY": "gsk_your_key",
"GH_TOKEN": "ghp_your_token"
}
}
}
}
Restart Claude Desktop. The following tools will be available: get_emails, compose_email, search_ai_ml, github_analyzer, tailor_resume_tool.
API Reference
All endpoints are available at http://localhost:5000 and accept/return JSON.
Health Check
GET /health
{ "status": "ok", "server": "waitress" }
Inbox Sorter
POST /tools/get_emails
{ "max_results": 10, "filter_priority": "all" }
Email Composer
POST /tools/compose_email
{
"to": "recipient@example.com",
"purpose": "Project status update",
"key_points": ["Milestone completed", "Next steps"],
"tone": "professional",
"auto_send": false
}
AI/ML Search
POST /tools/search_ai_ml
{ "query": "LoRA fine-tuning efficiency", "depth": "advanced", "max_results": 10 }
GitHub Analyzer
POST /tools/analyze_github
{ "action": "repo_overview", "ai_summary": true }
{ "action": "review_code", "repo": "hermes", "file_path": "main.py" }
Available actions: list_repos · repo_overview · commit_activity · readme_quality · stale_repos · review_code · tech_stack · audit_dependencies
Resume Tailor
POST /tools/tailor_resume
{
"role": "Senior ML Engineer",
"company": "Google DeepMind",
"job_description": "...",
"existing_resume": "...",
"mode": "full"
}
Available modes: full · quick · batch
Environment Variables
| Variable | Required | Description |
|---|---|---|
GROQ_API_KEY |
✅ | Groq API key — obtain from console.groq.com |
GH_TOKEN |
✅ | GitHub Personal Access Token with repo scope |
GMAIL_CREDENTIALS |
✅ | Path to credentials.json — defaults to project root |
PORT |
❌ | HTTP server port — defaults to 5000 |
GUNICORN_RELOAD |
❌ | Set to true to enable Gunicorn auto-reload on Linux |
Important: Use
GH_TOKENrather thanGITHUB_TOKEN. TheGITHUB_prefix is reserved by GitHub Actions and cannot be used as a custom secret name.
Development
# Run full test suite
venv\Scripts\python -m pytest tests/ -v
# Run with coverage report
venv\Scripts\python -m pytest tests/ -v --cov=tools --cov=services --cov-report=term-missing
# Lint
venv\Scripts\python -m flake8 tools/ services/ main.py serve.py --max-line-length=130
# Diagnose API key issues
python check_groq.py
Troubleshooting
ModuleNotFoundError: No module named 'fcntl'
Gunicorn does not support Windows. Use python serve.py or waitress-serve --port=5000 main:app instead.
Groq 401 Invalid API Key
The key is expired or revoked. Run python check_groq.py for a full diagnosis. Obtain a replacement key from console.groq.com/keys. Note that VS Code may cache stale .env values — running from a plain terminal window resolves this.
UI renders as raw CSS text
Perform a hard refresh with Ctrl+Shift+R. If the issue persists, confirm Flask is using send_from_directory(os.path.join(BASE_DIR, "ui"), "index.html") in main.py.
MCP tools not appearing in Claude Desktop
Verify that serve.py --mcp is specified in claude_desktop_config.json. Ensure all paths use double backslashes on Windows. Restart Claude Desktop after any configuration change.
VS Code terminal not loading .env
Add "python.terminal.useEnvFile": true to VS Code User Settings (JSON), or run from a plain cmd window outside VS Code.
Roadmap
- [ ] GitHub Actions monitor — workflow status and failure alerts
- [ ] Email-to-Issue bridge — create GitHub issues directly from emails
- [ ] Batch resume mode UI
- [ ] Export results as PDF
- [ ] Dark / light theme toggle
License
This project is licensed under the MIT License. See LICENSE for details.
<div align="center">
Built by rayyan666 · Powered by Groq llama-3.3-70b · Served by Waitress / Gunicorn · MCP via FastMCP
</div>
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.