Subscription Tracker

Run all of these searches (each with maxResults: 50):

1. "subscription renewal" OR "subscription confirmed" OR "your subscription" after:YYYY/M/D
2. "payment receipt" OR "payment confirmation" OR "payment processed" after:YYYY/M/D
3. "billing statement" OR "invoice" OR "your bill is ready" after:YYYY/M/D
4. "trial ending" OR "trial expires" OR "trial will end" OR "free trial" after:YYYY/M/D
5. "auto-renewal" OR "auto-renew" OR "will be charged" OR "upcoming charge" after:YYYY/M/D
6. "membership renewal" OR "plan renewal" OR "renews on" after:YYYY/M/D
7. subject:(receipt OR invoice OR payment OR renewal OR subscription) after:YYYY/M/D

Deduplicate results by messageId across all searches before proceeding.

False positive filtering: Before reading the full message, do a quick triage on the snippet and sender:

• Keep: Emails from billing/payment senders (stripe.com, razorpay.com, paypal.com, billing-noreply@, invoice+@, receipts+@, subscriptions@), or emails whose snippet contains currency amounts, receipt/invoice numbers, or payment confirmation language.
• Skip: Newsletter/Substack emails that merely contain the word "subscription" in their body (e.g., "You're reading the free edition of..." or "Your subscription to this newsletter"). Look at the sender domain — substack.com, every.to, beehiiv.com are almost always newsletters, not billing emails.
• Skip: Job alerts, HR tickets, password resets, and marketing emails that happen to contain billing keywords in unrelated context.

This filtering step is important because broad keyword searches will return many false positives from newsletters and marketing emails. Focus on emails where a financial transaction actually occurred or is about to occur.

Step 2: Read and Parse Emails

For each unique message found, use Gmail:gmail_read_message to get the full content. Important: Some billing emails (especially from Razorpay, Stripe) have rich data in the snippet field but a minimal plain-text body. Always check BOTH the snippet and body for data extraction — the snippet often contains the most useful structured information (amounts, dates, plan names) that may be stripped from the body. Extract the following fields from each email:

Additionally, for each subscription compute and store:

• days_until_renewal: integer — number of days from today to the next renewal date (negative if past due, null if unknown)
• urgency_tier: one of urgent (≤7 days), soon (8–30 days), upcoming (31–90 days), later (91–365 days), unknown — derived from days_until_renewal

Parsing guidance:

• If a field isn't found, set it to null — don't guess or hallucinate values
• For amounts, normalize to include currency symbol (e.g., "$9.99", "₹299", "€14.99")
• For Razorpay subscription emails, the snippet often contains the next due date, plan name, and amount — extract these directly. Look for "Next due on" and "Subscription Plan" fields.
• For Stripe receipt emails, the subject contains a receipt number (e.g., "#2360-9962-4668") and the snippet often contains "Amount paid" and "Date paid".
• For dates, normalize to ISO format YYYY-MM-DD
• Group multiple emails from the same service — keep the most recent/relevant one
• When the same service appears multiple times, consolidate: use the latest email's data but note the payment history

Step 3: Consolidate, Project Renewals, and Prioritize

After parsing all emails:

1. Group by service name — merge entries from the same service (e.g., multiple Spotify receipts → one Spotify entry)
2. Pick the most informative email per service — prefer renewal notices or upcoming billing alerts over past receipts
3. Calculate next renewal date — if only past payment dates are available and the billing cycle is known, project the next renewal date forward from the last payment date. Keep projecting until you find the next date that is in the future (relative to today). For annual subscriptions, this could be up to 12 months out.
4. Assign urgency tier based on the next renewal date (or cancellation deadline, whichever is sooner):

5. Sort by urgency, then by date — within each tier, sort by nearest renewal date first. The output should read like a timeline: what's due this week → this month → this quarter → later this year. One-time purchases and items with no projected renewal go to the very bottom.

Step 4: Generate the Output

Build the output as a React (.jsx) artifact saved to /mnt/user-data/outputs/subscription-tracker.jsx.

The artifact should include:

Header Section

• Title: "Subscription Tracker"
• Urgency banner: If any renewals are due within 7 days, show a prominent alert: "⚠️ X renewal(s) due this week"
• Summary stats: total monthly spend (estimate), number of active subscriptions, count of renewals in next 30 days
• Date range scanned

Interactive Table

The table should be organized into urgency sections with clear visual separation:

Section headers: "Due This Week", "Due This Month", "Next 3 Months", "Later This Year" — each section only appears if it has entries. Empty sections are hidden.

Columns (all sortable by clicking headers):

• Urgency indicator (color-coded dot matching tier: 🔴 red, 🟠 orange, 🟡 yellow, ⚪ gray)
• Service Name
• Amount
• Billing Cycle
• Days Until Renewal (e.g., "3 days", "2 weeks", "4 months" — human-readable relative time, with exact date on hover/tooltip)
• Plan/Tier
• Cancellation Deadline (highlight in red if within 7 days)
• Payment Link (clickable "Pay" button if available, otherwise "—")

Features

• Search/filter bar to filter by service name
• Urgency filter tabs: All / This Week / This Month / 3+ Months — default view is "All" but the visual weight is on near-term items
• Sort by any column (click header to toggle asc/desc) — default sort is by renewal date ascending (soonest first)
• Rows in the "Later This Year" section should appear with reduced opacity or muted colors so they don't compete for attention with urgent items
• The "Due This Week" section should have a subtle red/warm background to draw the eye
• Responsive layout

Design Direction

• Clean, dashboard-style aesthetic — not generic
• Use a warm color palette: soft backgrounds with sharp accent colors for status indicators
• Distinct typography: use a characterful sans-serif for headers (import from Google Fonts)
• Subtle card-based layout with hover states
• No purple gradients or generic AI look

Download Button

Include a "Download as Spreadsheet" button that triggers generation of the .xlsx file by calling sendPrompt('Generate the subscription spreadsheet for download').

Step 5: Generate the Spreadsheet

When the user clicks "Download as Spreadsheet" or asks for a spreadsheet export, create an .xlsx file using the script at scripts/generate_xlsx.py.

Run the script with the subscription data as JSON input:

echo '<JSON_DATA>' | python3 /path/to/subscription-tracker/scripts/generate_xlsx.py /mnt/user-data/outputs/subscription-renewals.xlsx

The script expects a JSON array of subscription objects piped to stdin and outputs a formatted .xlsx file.

Then present the file to the user with present_files.

Important Notes

• Privacy: All data stays within the conversation. Remind the user that their email content is being read to extract subscription info.
• Accuracy disclaimer: Tell the user that the extracted data is based on what's visible in their emails from the last 30 days, and some subscriptions may not appear if no related emails were received in that window.
• No hallucination: If a data field cannot be confidently extracted from the email, leave it as null/unknown rather than guessing. The artifact should show "—" for missing fields.
• Currency: Detect and preserve the original currency from the email. Don't convert currencies.
• One-time purchases: Some receipts may be for one-time purchases (workshops, courses, lifetime deals) rather than recurring subscriptions. If there's no billing cycle or renewal date, classify these as "one-time" and include them in a separate section at the bottom of the table. They're still useful for the user to see but shouldn't appear as "upcoming" renewals.
• Time sensitivity: When presenting dates, highlight anything due within the next 7 days prominently.