Stock PIlot

Updated todo list

Project Story

About the Project

StockPilot is an AI-powered trading research terminal that helps users explore markets, analyze stocks, review SEC filings, compare technical charts, and manage a virtual portfolio.

It combines real market data with WebMCP tools, allowing an AI agent to interact directly with the application. Users can ask the agent to research a ticker, open stock pages, analyze filings, draw chart patterns, screen stocks, and simulate trades.

Inspiration

We wanted to make financial research more interactive and accessible. Traditional market tools often require users to navigate through many separate screens, interpret complex charts, and manually connect information from news, filings, financial data, and portfolio performance.

StockPilot was inspired by the idea of an AI assistant that does more than answer questions. It should be able to take meaningful actions inside the trading terminal, such as opening an analysis page, highlighting important SEC passages, configuring a comparison chart, or creating a simulated position.

How We Built It

StockPilot was built with:

Next.js and React for the application interface
TypeScript for type-safe application logic
Finnhub for quotes, company profiles, news, recommendations, financials, earnings, and SEC filing metadata
Supabase for persisted market data and application storage
WebMCP for exposing StockPilot actions to compatible AI agents
Local OHLCV datasets for historical charting, indicators, pattern detection, and backtesting
Tailwind CSS and reusable UI components for the trading terminal interface

The application includes tools for:

Stock search and detailed quote lookup
Market news and sector performance
Stock screening and ranking
SEC filing review and risk analysis
Technical pattern detection
Support and resistance calculations
Portfolio analysis and risk scoring
Virtual buying and selling
Price alerts and watchlists
Strategy backtesting
Multi-stock chart comparison and technical drawings

Challenges We Ran Into

One major challenge was coordinating AI actions with a client-side application. Fetching data is straightforward, but actions such as navigation, chart updates, and filing highlights must be synchronized with the correct page and component state.

We also had to handle incomplete or unavailable provider data. SEC filings may not always include a readable report URL, and free-tier APIs can return limited results. To keep the experience reliable, StockPilot uses fallbacks, validation, normalized filing content, and best-effort analysis.

Another challenge was making technical chart interactions precise. Drawing points need to align with real trading dates and meaningful price ranges, so the application snaps dates to available candles and clamps prices to valid OHLCV boundaries.

Accomplishments We’re Proud Of

We are proud that StockPilot is more than a dashboard. It is an interactive environment where an AI agent can:

Navigate users to relevant pages
Retrieve and summarize real market data
Analyze SEC filings and calculate risk signals
Highlight important filing passages
Configure chart styles, indicators, and time ranges
Detect and draw technical patterns
Compare multiple companies
Simulate portfolio actions while enforcing authentication boundaries

The WebMCP integration gives the AI a practical interface to the product instead of limiting it to text-only responses.

What We Learned

We learned that building an AI-enabled application requires designing both the user interface and the action interface at the same time. Tool descriptions, input schemas, validation, navigation behavior, and error messages are all part of the user experience.

We also learned the importance of grounding financial analysis in concrete data. Historical prices, filing documents, recommendation trends, and portfolio state should be traceable to actual application data rather than invented assumptions.

Finally, we learned that asynchronous workflows need careful handling. Data fetching, page navigation, event dispatching, and component mounting must work together so that an AI-triggered action remains understandable and reliable to the user.

What’s Next for StockPilot

Next, we plan to improve StockPilot with:

More reliable cross-page AI action handoffs
Expanded market and sector coverage
More advanced portfolio analytics
Additional technical indicators and drawing tools
Richer SEC filing comparisons across reporting periods
Persistent user portfolios and watchlists through Supabase
More robust authentication and account management
Improved backtesting with transaction costs and risk controls
Broader WebMCP support for agent-driven financial workflows

Our long-term goal is to make StockPilot a trustworthy collaborative workspace where people and AI agents can research markets together while keeping decisions transparent, data-grounded, and user-controlled.