Ross MCP
Personal life admin MCP server that manages Apple Reminders from any Claude session via a cloud relay to a local Mac agent.
README
Ross MCP
Personal life admin system — manage Apple Reminders, Outlook email/calendar, and Apple Notes from Claude, ChatGPT, or any MCP/API client.
Architecture: Client (Claude/ChatGPT/API) → Cloud Relay (Fly.io) → Local Mac Agent → Apple APIs / Microsoft Graph
Features
| Category | Tools |
|---|---|
| Apple Reminders | Create, list, complete reminders |
| Outlook Email | Search, read, draft, send, schedule, archive emails |
| Outlook Calendar | List events, create/update/cancel events, find free slots |
| Apple Notes | Search, read, create notes, list folders |
| Voice Memos | List recordings, transcribe with speaker diarization (Deepgram) |
24 tools accessible from:
- Claude (web/desktop/CLI) via MCP protocol
- ChatGPT via Custom GPT Actions (OpenAPI)
- Any HTTP client via REST API
Quick Start
1. Clone and set up Python
cd ross-mcp
python3 -m venv .venv
source .venv/bin/activate
pip install -r agent/requirements.txt
pip install mcp httpx
2. Configure environment
cp .env.example .env
# Edit .env — set RELAY_API_KEY (must match the Fly.io secret)
3. Set up Outlook (one-time)
# Install Azure CLI
brew install azure-cli
# Log in to Azure
az login
# Register the app and save credentials to .env
./agent/setup_azure.sh
# Run the OAuth login (opens browser)
python3 -c "
import asyncio
from dotenv import load_dotenv
load_dotenv()
from agent.services.outlook_auth import OutlookAuth
auth = OutlookAuth()
asyncio.run(auth.authorize())
print('Success!' if auth.is_authenticated else 'Failed')
"
You only need to do this once per Mac. The refresh token auto-renews every 3 days.
4. Run the agent
source .venv/bin/activate
python -m agent.agent
The agent will:
- Connect to the cloud relay via WebSocket
- Start a local web UI at http://127.0.0.1:8001
- Listen for commands from any client
5. Install as auto-start service
./agent/install.sh
Creates a launchd service that starts on boot and keeps running.
Connecting Clients
Claude Web / Desktop (MCP)
Connect to the remote MCP endpoint:
| Setting | Value |
|---|---|
| URL | https://ross-mcp-relay.fly.dev/mcp/mcp |
| Transport | Streamable HTTP |
| Auth | Bearer token (your RELAY_API_KEY) |
Claude Code (CLI)
Add to ~/.claude/settings.json:
{
"mcpServers": {
"ross-life-admin": {
"command": "/path/to/ross-mcp/.venv/bin/python",
"args": ["/path/to/ross-mcp/mcp_server.py"],
"env": {
"RELAY_API_KEY": "your-api-key",
"MCP_RELAY_URL": "https://ross-mcp-relay.fly.dev"
}
}
}
}
ChatGPT (Custom GPT)
- Create a Custom GPT at chat.openai.com
- Add an Action → Import from URL:
https://ross-mcp-relay.fly.dev/openapi.json - Set auth to Bearer with your
RELAY_API_KEY
This imports all 22 tool endpoints. See the Swagger UI at https://ross-mcp-relay.fly.dev/docs.
Direct REST API
curl -X POST https://ross-mcp-relay.fly.dev/api/command \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"type": "create_reminder", "payload": {"title": "Buy milk"}}'
Deployment
Cloud Relay (Fly.io)
The relay runs on Fly.io in the lhr region. To deploy changes:
fly deploy
Manage secrets:
fly secrets list --app ross-mcp-relay
fly secrets set RELAY_API_KEY=your-key --app ross-mcp-relay
View logs:
fly logs --app ross-mcp-relay
Local Agent
Manual start:
source .venv/bin/activate
python -m agent.agent
Launchd service commands:
| Action | Command |
|---|---|
| Install | ./agent/install.sh |
| Stop | launchctl unload ~/Library/LaunchAgents/com.ross.mcp-agent.plist |
| Start | launchctl load ~/Library/LaunchAgents/com.ross.mcp-agent.plist |
| Restart | Unload then load |
| Logs | tail -f ~/Library/Logs/mcp-agent/mcp-agent.log |
| Errors | tail -f ~/Library/Logs/mcp-agent/mcp-agent.err |
Setting up a second Mac
- Clone the repo
- Set up venv and install dependencies
- Copy
.envand changeAGENT_NAMEto identify the machine - Run
./agent/setup_azure.shthen the OAuth login (one-time) - Run
./agent/install.sh
Both agents connect to the same relay — commands route to whichever is online.
Voice Memo Transcription
Record meetings on iPad using Voice Memos, share to iCloud Drive, then transcribe via Claude or ChatGPT.
Setup
- Create a "Meeting Recordings" folder in iCloud Drive (done automatically by the agent)
- Create an iOS Shortcut called "Save Meeting Recording" on your iPad:
- Action: Save File → iCloud Drive / Meeting Recordings (Ask Where to Save: off)
- Enable Show in Share Sheet, set receives to Audio
- Add your
DEEPGRAM_API_KEYto.env
After a meeting
- Open Voice Memos on iPad → tap ... → Share → Save Meeting Recording
- Tell Claude: "Transcribe my meeting with [name] from this morning"
- Claude finds the recording, transcribes with speaker diarization, enriches with summary and action points, and saves as an Apple Note
Remote Endpoints
| Endpoint | URL | Auth |
|---|---|---|
| Dashboard | https://ross-mcp-relay.fly.dev/ |
API key in UI |
| MCP (Claude) | POST https://ross-mcp-relay.fly.dev/mcp/mcp |
Bearer token |
| REST API | POST https://ross-mcp-relay.fly.dev/api/command |
Bearer token |
| Tool endpoints (ChatGPT) | POST https://ross-mcp-relay.fly.dev/api/tools/* |
Bearer token |
| Swagger UI | https://ross-mcp-relay.fly.dev/docs |
None (read-only) |
| Status | GET https://ross-mcp-relay.fly.dev/api/status |
Bearer token |
| Agent WebSocket | wss://ross-mcp-relay.fly.dev/ws/agent |
Bearer token |
Secrets
| Secret | Location | Notes |
|---|---|---|
RELAY_API_KEY |
.env (local) |
Shared by agent, MCP server, and clients |
RELAY_API_KEY |
Fly.io secrets | Set via fly secrets set |
MS_CLIENT_ID / MS_CLIENT_SECRET |
.env (local) |
Azure AD app credentials |
.outlook_tokens.json |
Project root (gitignored) | OAuth tokens, auto-refreshed |
DEEPGRAM_API_KEY |
.env (local) |
For voice memo transcription |
Regenerate API key:
python3 -c "import secrets; print(secrets.token_urlsafe(32))"
# Update: .env, fly secrets set, and any client configs
Project Structure
ross-mcp/
├── agent/ # Local Mac agent
│ ├── agent.py # Main agent (WebSocket client + command dispatch)
│ ├── web.py # Local web UI (port 8001)
│ ├── setup_azure.sh # Azure AD app registration script
│ ├── install.sh # Launchd auto-start installer
│ └── services/
│ ├── reminders.py # Apple Reminders via EventKit
│ ├── notes.py # Apple Notes via AppleScript
│ ├── outlook_auth.py # OAuth2 for Microsoft Graph
│ ├── outlook_mail.py # Outlook email operations
│ ├── outlook_calendar.py # Outlook calendar operations
│ └── voice_memos.py # Voice memo transcription (Deepgram)
├── relay/ # Cloud relay (Fly.io)
│ ├── relay.py # FastAPI hub (WebSocket + HTTP + dashboard)
│ ├── mcp_endpoint.py # Remote MCP server (streamable-http)
│ ├── openai_endpoints.py # REST endpoints for ChatGPT Actions
│ ├── Dockerfile
│ └── requirements.txt
├── shared/
│ └── messages.py # Command/Response schemas (25 command types)
├── mcp_server.py # Local MCP server (stdio, for Claude Code)
├── fly.toml # Fly.io config
├── .env.example # Environment template
├── ROADMAP.md # Planned features
└── SPRINT.md # Sprint log
Links
- Roadmap — Planned features
- Sprint Log — Development history
- Fly.io Dashboard — Deployment management
- Swagger UI — API documentation
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.