Skip to main content
An AI assistant can help an SMB owner turn raw bank activity into categorized accounting data. This is useful for routine bookkeeping workflows where the owner wants to review unresolved transactions, understand what needs attention, and approve reasonable categorizations.

​
Questions And Uncertainties

Business owners and bookkeepers often want help answering:
  • Which bank transactions still need review?
  • Which transactions have obvious categories?
  • Which transactions might be personal, duplicated, or excluded from the books?
  • Which transactions should be matched to existing business activity instead of categorized as standalone income or expense?
  • Which proposed categorizations are safe to apply in bulk?

​
Prompts And Tools

Users can ask prompts like:
  • “Show me the transactions that still need review.” This typically triggers get_bank_transactions.
  • “Suggest categories for these uncategorized transactions.” This typically triggers get_bank_transactions and get_chart_of_accounts.
  • “Categorize this hardware store transaction as supplies.” This typically triggers get_chart_of_accounts and categorize_bank_transaction.
  • “Apply these approved categories in bulk.” This typically triggers bulk_match_or_categorize_bank_transactions.
  • “Match these deposits to their suggested matches.” This typically triggers get_bank_transactions and match_bank_transaction or bulk_match_or_categorize_bank_transactions.

​
Expected Interactions

The agent should first gather the relevant transactions and accounting context, then present a proposed set of actions for review. For ambiguous transactions, the agent can ask follow-up questions before changing the books. Once the user approves the plan, the agent can submit single or bulk categorization and matching requests through the MCP tools.

​
Details And Variations

Splits are useful when one bank transaction belongs to multiple accounting categories. The agent can propose split entries, but amounts must sum to the transaction amount before the categorization can be submitted. See Split Categorizations for the API request shape. Bank transaction matches are useful when the transaction should be reconciled to existing activity instead of categorized directly. The agent can inspect suggested matches returned with transaction data and confirm the right match after user approval. See Match Bank Transaction for the API behavior. Bulk operations are useful after the user reviews a batch of proposed actions. Internally, the MCP server triggers the same Layer API workflows documented in the bank transaction API reference, including Bulk Match or Categorize Bank Transactions.