davepoon/tidy

Tidy Up Uncategorized Transactions

Batch-categorize uncategorized transactions by clustering similar ones and applying categories in bulk.

Workflow

1. Fetch uncategorized transactions. Call the query MCP tool:

   { "detail": true, "is_uncategorized": true, "period": "last_90d", "limit": 200, "sort": "-amount" }
   

   If $ARGUMENTS contains a time period (e.g. "this month", "last 30 days"), use that instead of last_90d.

2. Research unknown transactions. For transactions you can't identify from the description alone:

   • Web search first (if available): Search for the merchant name, any phone numbers or domains in the description, or the raw description itself. This often reveals the business behind cryptic processor names.
   • Search the user's email (if available): Search for the party/merchant name to find order confirmations or receipts. If that doesn't match, search for the exact dollar amount (e.g. "$47.23") to find receipts that way.

3. Cluster by pattern. Group the results by normalized description or party name. For each cluster, note the count and total amount.

4. Suggest categorization. For each cluster, propose:

   • A category (pick from the user's existing categories)
   • A party name (the clean merchant/counterparty name)

5. Present to the user. Show a table or list of clusters with:

   • Pattern / merchant name
   • Count of transactions
   • Total amount
   • Suggested category
   • Whether you recommend creating a rule

   Ask the user to approve, modify, or skip each cluster.

6. Prefer rules over one-off annotations. If a cluster has more than one transaction, or the merchant is likely to appear again (subscriptions, regular stores, utilities, etc.), create a rule rather than annotating individual transactions. Rules automatically categorize future transactions too.

   • Preview first: admin { "entity": "rule", "action": "preview", ... }
   • Show the preview (how many existing transactions would match)
   • If user confirms, create: admin { "entity": "rule", "action": "create", ... }

7. Annotate the rest. For truly one-off transactions where a rule wouldn't help, apply directly:

   { "action": "categorize", "filter": { "search": "<pattern>" }, "category_name": "<approved_category>" }
   

   Also set the party if one was approved:

   { "action": "set_party", "filter": { "search": "<pattern>" }, "party_name": "<approved_party>" }
   

8. Summarize. Report how many transactions were categorized, how many rules were created, and how many uncategorized transactions remain.

Tone

Stick to the facts. Present findings and suggestions without judgement — no commentary on spending habits. Just clear, plain-language observations and actionable options.