Build your own finance tools with real bank data in Cursor
Add the BankSync MCP server to Cursor and query transactions, balances, investments and loans from 12,000+ banks while you build.
src/spend-by-category.ts
import { transactions } from './data' // From get_transactions, last 90 daysexport function spendByCategory() { const totals = new Map<string, number>() for (const tx of transactions) { if (tx.amount >= 0) continue const key = tx.category ?? 'Other' const prev = totals.get(key) ?? 0 totals.set(key, prev - tx.amount) } return [...totals].sort(byAmount)}
Agent
Connect Cursor in three steps.
- 01
Create a scoped key
Connect your banks in BankSync, then create an API key with read scopes under Developers.
New API key
NameMCP clientScopesbanks:readfeeds:readKeybsk_••••••••3f9a - 02
Add the server
Put the BankSync entry in .cursor/mcp.json in your project, or ~/.cursor/mcp.json for every project.
.cursor/mcp.json
{ "mcpServers": { "banksync": { "url": "https://mcp.banksync.io", "headers": { "X-API-Key": "bsk_your_key_here" } } } } - 03
Use it in Agent mode
Ask the agent for a lookup or a whole feature; it reads the rows it needs.
Pull the last 90 days of transactions and write a spend-by-category helper with a test.
get_transactions · 412 rows · Business Checking ••4021
Done. spendByCategory() is in src/, and the test uses 20 real rows as fixtures. Software is the top category at $4,212.
What you can build.
- Finance tool development
- Build financial applications with real bank data schemas, transaction types, and account structures available directly in your editor.
- Financial data queries
- Query live transactions, balances, and holdings from Cursor's Composer mode to understand data shapes and build accurate integrations.
- Code generation with real schemas
- Generate TypeScript types, API clients, and data models based on actual bank data structures from 12,000+ institutions.
- Budget app scaffolding
- Scaffold complete budget and expense tracking applications with Cursor Agent mode using real financial data as reference.
- API client generation
- Auto-generate typed API clients for bank data endpoints with full request/response schemas from BankSync MCP tools.
- Test fixture creation
- Create realistic test fixtures and mock data based on actual bank transaction formats, account types, and investment structures.
Change assistants. Keep your data.
What Cursor gets.
- Server
- Remote and hosted at https://mcp.banksync.io over Streamable HTTP. Nothing to install.
- Tools
- 227 documented tools: bank data, feeds and jobs, tables, dashboards, automations and enrichments. Tools reference
- Sign-in
- A scoped API key, sent in the X-API-Key header.
- Config
- .cursor/mcp.json in the project root, or ~/.cursor/mcp.json globally. Project config wins.
- Access
- Set by the scopes you grant. Read scopes keep the connection read-only; no tool can move money.
- Rate limits
- 20 tool calls a minute on Standard, 40 on Professional, 60 on Business, shared across a workspace.
- Banks
- 12,000+ institutions in the US, Canada, the UK, Europe and Australia, including Consumer Data Right connections.
Questions, answered.
How do I connect Cursor to my bank data?
Generate a BankSync API key, then add the BankSync MCP server configuration to your .cursor/mcp.json file. Cursor will automatically detect the server and make 227 financial data tools available in Composer and Agent modes.
What bank data can I access in Cursor?
You can access transactions, balances, investment holdings, trades, and loan data from 12,000+ banks through the BankSync MCP server directly in Cursor.
Does Cursor support MCP servers like BankSync?
Yes. Cursor supports MCP through manual configuration in .cursor/mcp.json. It has the strongest tool execution reliability among AI coding tools and works seamlessly with BankSync MCP.
Is the BankSync MCP connection secure in Cursor?
Yes. Authentication uses scoped API keys with granular permissions. All communication is encrypted, and you can revoke access anytime from the BankSync dashboard.