What's new in Zenny — April 28, 2026

This release makes Zenny more useful for people who want to ask deeper questions about their money from tools like Cursor, Claude, and other MCP clients. The new connection model gives those tools a flexible, read-only way to analyze your Zenny data while keeping account ownership and access limits enforced by the server.

More flexible financial questions

Connected AI clients can now ask broader questions across transactions, receipts, accounts, budgets, goals, debts, investments, and alerts without Zenny needing a separate tool for every possible question. That means you can ask things like “what did I spend on car expenses in 2025?” or “which merchants drove the biggest increase last quarter?” and get a direct answer from your own scoped data.

The query interface is designed for summaries first, so answers favor totals, counts, trends, and grouped insights before drilling into individual records. When you do need the underlying rows, Zenny applies stricter caps so responses stay focused and manageable.

Safer access for external tools

API keys now include an explicit mcp:sql:read scope for read-only AI access. Zenny checks that scope before allowing MCP queries, and every query runs through server-owned, user-scoped relations rather than raw application tables.

This keeps tenant isolation as the default behavior: connected tools can analyze your financial data, but they only see the data tied to your account and only through the approved read-only surface.

Cleaner voice answers

Voice answers now stay centered on the transcript, status, spoken summary, and any errors that need attention. Navigation cards, open-page shortcuts, and suggested next steps have been removed from the answer overlay so the experience feels calmer and more focused while you are asking questions out loud.

Try connecting a local MCP client or using voice to ask a focused question about a specific category, merchant, or date range.