list_unusual_transactions

Flags expense transactions that look unusual compared with the user's own history. Read-only. Amounts are in currency major units, not cents.

Use with an AI agent

Export this page as Markdown or bundle it with a task for your agent.

Try a task
Preview the prompt that will be copied

Access

  • Effect: Read-only
  • Required scope: read (the mcp scope also grants it)

Example request

The JSON-RPC message an MCP client sends to call this tool. Your client builds it for you; this is for reference and debugging in a tool such as MCP Inspector.

json
{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_unusual_transactions",
    "arguments": {}
  }
}

Example IDs refer to fictional demo data. Replace them with IDs from your account before calling the real server.

Description

Flags expense transactions that look unusual compared with the user's own history. Read-only. Amounts are in currency major units, not cents.

Use this for: "Is that transaction unusual?", "Anything odd in my spending this month?", "Did I get charged twice?", "Any surprising charges lately?".

Each flagged transaction has a severity (high or medium) and reasons: `amount_high_for_vendor` (far above what that vendor usually charges), `amount_high_for_category` (far above the usual for that category), `new_vendor_large_amount` (first charge from a vendor and larger than 95% of earlier expenses), and `possible_duplicate` (same vendor and amount within 48 hours). Each charge is compared only with charges before it, using up to a year of earlier history. Transfers are ignored.

A journal is compared once using its net expense category postings, not once per category or account line. Refund-only and net-zero journals are ignored.

from/to are RFC 3339 timestamps for the window to scan; the default is the last 30 days and the maximum is 365 days. limit defaults to 20 (max 100); results are ordered high severity first, then newest. accountIds is optional; call list_accounts for IDs. To check one specific charge, scan a window around its date and look for its transactionId.

Input parameters

NameTypeRequiredDescription
accountIdsarrayNo
fromstringNo
limitintegerNo
tostringNo
Full input schema (JSON Schema)
json
{
  "additionalProperties": false,
  "properties": {
    "accountIds": {
      "items": {
        "type": "string"
      },
      "type": [
        "null",
        "array"
      ]
    },
    "from": {
      "type": [
        "null",
        "string"
      ]
    },
    "limit": {
      "type": [
        "null",
        "integer"
      ]
    },
    "to": {
      "type": [
        "null",
        "string"
      ]
    }
  },
  "type": "object"
}