mathan_karthik

it will always free. i will not charge for anything.

I faced an unexpected fund shortage 3 months ago and had no idea what my net worth was. So I built a tool to solve it (for free).

Hey HY

Three months ago, I hit a low point. I needed to move some funds for an emergency and realised I had no idea where I actually stood.

My stocks were in Groww, my mutual funds in ICICI, I had physical gold in a locker, and I was juggling active loans. Every time I wanted an answer, I had to open four different apps and a physical diary.

I started tracking everything in an Excel sheet, but it was tedious. My wife also wanted to track our family finances, and I wanted to sharpen my AI knowledge. So, I decided to turn this into a weekend project.

The "3-Week" Evolution It started simple:

First Draft: Connected a real-time stock and MF tracker. It worked! Gold & Net Worth: Added a gold tracker and a dashboard to calculate my total net worth to the rupee. Collaboration: I added "Team Access" (Crew feature) so my wife and I could both view and update our shared corpus. My friends saw it, loved it, and started using it too. By week 3, it was a solid tool, but there was one massive pain point: Manual Entry.

The "Share Receipt" Spark Everything worked well, but logging every expense was still a manual entry drag. Since most of our spending happens via UPI, I was looking for a way to simplify it.

A friend suggested: "All these UPI apps have a 'Share' button after you pay. Just use that!"

That hit a spark. I created a UPI tracker where, after every transaction, I just share the receipt to my project's dedicated email. The system automatically reads it and enters the value for me. No more manual entry!

AI Helper & MCP Connector Since I wanted to use this to learn more about AI, I didn't stop at simple tracking.

MCP Connector: I've added a Model Context Protocol (MCP) connector. This means you can actually connect your financial data context to Claude or ChatGPT. You can ask your favourite LLM: "Based on my Corpus Tracker data, can I afford to buy a new car this year?" In-App Assistant: I’ve built a financial helper directly into the app chat. It helps you categorize, analyze, and process everything in natural language. What’s Inside? Unified Dashboard: Stocks (Groww-style), MFs, Gold, and Cash. Loan Manager: Track principal vs interest for EMIs. Share-to-Track: Forward UPI receipts via email/share for auto-logging. Crew/Family Access: Manage a shared corpus with your partner. AI Financial Helper: Chat with your data. MCP Enabled: Use your data context in Claude/ChatGPT. It's 100% Free I’m launching this for free to everyone. No hidden charges, no "premium" tiers for AI features—everything is open and available. I built this to solve my own problem, and I’m happy to share it with the community.

[Link to your app]

Feel free to ask me anything about the tech, the UPI parsing, or the AI integration. I'm actively looking for feedback to make it better!

P.S. — Special thanks to my wife for being the first alpha tester and to the friend who suggested the share button!

I'm building a personal finance tracker for Indian users and thinking about adding a feature where users can share their UPI payment receipts (screenshots from PhonePe, GPay, Paytm, etc.) directly into the app.

How it would work:

After a UPI payment, the user shares the receipt screenshot to the app The app parses the receipt (OCR/image parsing) to extract amount, merchant, and date It automatically creates a spending entry If the merchant is unrecognized, on next login the user is prompted to categorize it Once mapped, future payments to the same merchant are auto-categorized Why receipts instead of email/SMS parsing:

UPI apps (PhonePe, GPay, Paytm) all have a "Share" button on every receipt No need for SMS permissions or email forwarding setup Works cross-platform — the receipt format is consistent enough for OCR Less privacy-invasive than reading SMS or email My questions for HN:

Would you use this? Is the friction of "share receipt" low enough? Any existing apps doing this well (India or globally)? Is OCR reliable enough for this, or is there a better parsing approach?

Fair point on the privacy policy. Since this handles financial data (net worth tracking), having a clear, accessible privacy policy would build trust—especially if accepting tips/payments. Even for a free tool, explicit data handling practices (what's stored, how long, where) matter a lot to users.

Would be great to see this addressed in the next update.

I’m open to your suggestions. Please guide me, as even a small piece of feedback from you could help me solve this problem.

I'm an Indian developer. I built Corpus Tracker for myself because nothing else fit how I track money: monthly budget by category, stocks, mutual funds, gold, and emergency fund in one place, with INR and support for how we actually save and invest. What it does: Budget – Income/expense by month, categories, copy previous month Stocks – Indian equities (NSE), Zerodha/Groww CSV import Mutual funds – Holdings and transactions, AMC logos Gold – Holdings in grams, live price, simple P&L Emergency fund – Goal and contributions Multi-profile – Separate “corpuses” (e.g. family, side hustle), invite members with roles I'm the only one building it. It’s tailored to my own workflow and has become my default for monthly budget tracking and net worth. No premium paywall for core features.

Thanks for the thoughtful questions and for sharing your experience!

On country-specific tax logic and edge cases: I’m planning to design the core with extensibility in mind—so things like tax rules, invoice sequencing, and regional compliance can be handled via plugins or configuration modules. For example, German invoice sequencing or Singapore GST could be implemented as optional adapters, and I’d love help identifying the most common patterns and pain points.

Authentication: Unified auth/session handling is a big goal, but I recognize each platform’s OAuth flow is a bit different. My idea is to abstract the common flows and provide helper functions, but also allow for custom overrides where platforms deviate too much. I’d be interested in hearing more about where OAuth gets especially tricky for your use cases.

Data mapping/sync: This is definitely the hardest part. My current thinking is to define a canonical data model (for invoices, customers, etc.), then provide mapping layers for each platform. There will be tools to help with syncing and conflict resolution, but I expect this will need ongoing refinement as APIs evolve and edge cases pop up.

Would love to keep this conversation going—if you have specific war stories or must-have features, please let me know! And if anyone’s interested in collaborating, I’d be happy to open up the repo early for feedback and contributions.

agree even langgraph has may tool for web search, so now it's like school project.