Milestone 3 — AI Agent with MCP

Weekly review pulse agent: ingest App Store / Play Store exports → synthesize themes and actions → publish to Google Docs and Gmail draft via Google Workspace MCP (no direct Google APIs).

Repository: github.com/alekhya2395/Milestone-3---AI-agent-MCP

Settings

Local setup from scratch (Windows / PowerShell):

# 1. Clone
git clone https://github.com/alekhya2395/Milestone-3---AI-agent-MCP.git
cd Milestone-3---AI-agent-MCP

# 2. Virtual environment
python -m venv .venv
.\.venv\Scripts\Activate.ps1

# 3. Install dependencies
pip install -r requirements.txt

# 4. Environment variables
copy .env.example .env
# Edit .env with your keys (Google OAuth, Alpha Vantage, etc.)

# 5. MCP setup (Cursor)
# Follow: phases/phase-01-mcp-setup/gcp-setup-checklist.md
# Then:  phases/phase-01-mcp-setup/runbook.md

# 6. Download reviews (Phase 2)
python phases/phase-02-review-ingestion/scripts/fetch-reviews.py --weeks 10
python phases/phase-02-review-ingestion/scripts/normalize-reviews.py

Step 5 details:

• GCP setup checklist — Google Cloud project, OAuth, APIs
• MCP runbook — connect google-drive and google-gmail in Settings → Tools & MCP

After step 6, review outputs are in data/raw/ (CSV) and data/reviews/reviews.json (filtered JSON).

Documentation

Doc	Purpose
docs/problemstatement.md	Goals and constraints
docs/architecture.md	System design
docs/implementationplan.md	Phase-wise plan
docs/eval.md	Milestone evaluation tracker
docs/decision.md	Architecture decisions
docs/deployment-railway.md	Deploy custom MCP server on Railway

Railway deployment

Deploy the weekly-pulse MCP server (review tools) to Railway:

railway up

Full guide: docs/deployment-railway.md

Google Gmail/Drive MCP remain on Google's servers — connect both from Cursor.

Project layout

├── .cursor/mcp.json          # Drive + Gmail MCP (secrets via env vars)
├── data/raw/                 # Store review exports
├── data/processed/           # Normalized reviews & pulse artifacts
├── docs/                     # Specs and per-phase eval.md
├── phases/                   # Phase work folders (runbooks, artifacts)
└── prompts/                  # Agent prompts (Phase 3+)

Current phase: 1 — MCP & foundation

1. Complete GCP setup checklist
2. Set env vars from .env.example
3. Restart Cursor → Settings → Tools & MCP → Connect google-drive and google-gmail
4. Run smoke tests per runbook
5. Sign off Phase 1 eval

MCP config location

• Project: .cursor/mcp.json (committed, no secrets)
• Credentials: GOOGLE_OAUTH_CLIENT_ID and GOOGLE_OAUTH_CLIENT_SECRET in .env or system environment

References

• Configure Google Workspace MCP servers
• Cursor MCP docs