Building a Personal Finance Dashboard for Under $20 with Claude Opus 5.5

One Developer's Budget-Friendly Monarch Alternative

On Reddit, a developer shared his project SharkFin — a personal finance dashboard inspired by the commercial app Monarch. What makes the project remarkable is the cost: he built a fully functional, highly customizable finance system for under $20.

That price isn't an exaggeration. The developer used a Claude Pro subscription ($20/month) and leveraged Opus 5.5's medium reasoning tier to build the frontend. By his account, the entire React frontend consumed only 35% of his weekly usage quota — leaving plenty of headroom for other work. For anyone looking to escape subscription-based finance software, it's a genuinely eye-opening approach.

Technical Architecture: A Stack of Open-Source Tools

SharkFin wasn't built from scratch. Instead, it cleverly strings together several mature open-source and third-party services:

• SimpleFin: Connects to bank accounts and pulls transaction data
• Actual Budget: Serves as the core budgeting engine and data storage layer
• React: A custom frontend interface generated by Opus 5.5

The most critical choice here is Actual Budget. The author specifically highlighted that Actual is open-source, meaning all financial data is stored locally — users retain complete ownership of their data. This stands in sharp contrast to cloud-based commercial products like Monarch or Mint, where your sensitive financial data always lives on the provider's servers.

For privacy-conscious users, this layered architecture — bank data ingestion (SimpleFin) + local budgeting engine (Actual) + custom frontend — offers a solution that balances convenience with data sovereignty.

SimpleFin is an open protocol and service designed for personal finance aggregation. It lets users securely pull transaction records from bank accounts via a standardized API without sharing banking credentials with third-party apps. Its core design philosophy is "authorize once, use everywhere" — users generate an access token on the bank's end, and budgeting apps use that token to read data rather than holding passwords directly. Compared to commercial aggregators like Plaid, SimpleFin prioritizes transparency and minimal permissions, and is developer-friendly enough for self-hosted integrations.

Actual Budget is a local-first, open-source budgeting app that stores data on the user's local device or self-hosted server by default, with end-to-end encrypted sync support. It was once a commercial product before going open-source in 2022, and provides a complete budgeting management API — making it ideal as a data backend for custom frontends. For users sensitive about financial data, this local-storage architecture means data won't be lost or monetized even if the service provider shuts down.

The Real-World AI-Assisted Development Experience

The author gave quite positive marks to Opus 5.5's medium reasoning tier. He was fully satisfied with the medium tier's output and never felt the need to bump up the reasoning intensity. This detail is worth noting for developers: in many real-world coding tasks, the highest reasoning level simply isn't necessary — a medium tier often strikes the right balance between cost and quality.

Even more interesting is his workflow. He had Opus 5.5 first research existing finance dashboards, pull reference screenshots, and then build the UI based on those visual references. This pattern of "let the AI research first, then implement" is becoming an increasingly common and effective paradigm in AI-assisted development.

Opus 5.5's reasoning tier mechanism is worth explaining. Claude's extended thinking feature allows the model to perform internal reasoning before delivering a final response, divided into different intensity levels (e.g., low, medium, high). The higher the tier, the more "thinking tokens" the model invests — and the more quota it consumes. The medium tier means the model does a moderate amount of internal reasoning, suitable for tasks like code generation and component design that have some complexity but don't require deep logical derivation. The author's choice of medium over the highest tier — with satisfying results — validates a practical principle: reasoning intensity should match task complexity. Blindly selecting the highest tier wastes quota and can sometimes introduce unnecessary complexity through over-analysis.

Core Methodology: Nail One Module First

The author's most important piece of advice is about development cadence:

Focus on making one part of the app really work well, then ride that momentum to keep expanding.

His own approach was to start with the cash flow page, polish it thoroughly, and then gradually add other pages. This "single breakthrough point, snowball expansion" strategy is especially effective in AI-assisted development — once the first module's code structure, component conventions, and data flow are clear and stable, subsequent modules can reuse a large number of established patterns, and the AI can generate code more accurately based on existing context.

Customizability: Separate Pages for Multiple Properties

What the author appreciates most about SharkFin is its high degree of customizability. As a landlord with multiple rental properties, he used Opus 5.5 to easily build dedicated monitoring pages for each property — letting him view monthly cash flow and top expenses for every unit individually.

One clever design choice: he only feeds net cash flow (or loss) into the main budget, rather than dumping every property's line items into it — avoiding clutter that would pollute the main budget's clarity. This kind of fine-grained business logic customization is exactly what general-purpose commercial finance software struggles to deliver, and precisely where self-built solutions shine.

What the Disappearing Barrier Means

The author closes with a reflection: "It's just wild how easy it is to build something like this now." That observation may be the most thought-provoking part of the whole case study.

A few years ago, building a system that connects to banks, integrates a budgeting engine, looks polished, and is fully customizable would have required substantial frontend, backend, and data integration work. Today, with mature open-source components and AI coding assistants, a single developer can pull it off over a weekend for the price of a cup of coffee.

The author has no plans to sell SharkFin, but says he'd consider open-sourcing it if there's enough interest. Regardless of whether it goes open-source, the case itself clearly demonstrates: AI is turning "building your own software" from a privilege reserved for technical experts into an increasingly accessible option for a much broader group of developers.

This trend is known in software engineering as the acceleration of the "citizen developer" phenomenon — people without formal programming backgrounds building software that meets their own needs using low-code, no-code, or AI-assisted tools. What makes SharkFin notable is that it's not a simple form-builder exercise — it's a genuine engineering project involving API integration, state management, and multi-source data aggregation. This signals that AI coding assistants are significantly raising the ceiling on "how complex a piece of software an individual can build." The long-term industry implication: many vertical SaaS products that relied on a "technical moat" to sustain subscription revenue — especially tools whose core value lies in data integration and visualization rather than complex algorithms — will face growing competitive pressure from user-built alternatives.