ezBookkeeping

Deploying a household finance server

Start with Docker or a binary on a home server, NAS or small VPS, and choose MySQL or PostgreSQL from day one, because moving away from SQLite later needs a third-party tool such as DBeaver. Set secret_key, generated with ezbookkeeping security gen-secret-key, set domain, and set root_url when a reverse proxy changes the scheme or port. Mount data, storage and log directories writable by UID 1000. Then register the first user and confirm that pictures and avatars load through the public address. The configuration reference lists every option and its environment variable.

Migrating history from another finance tool

Export from the old tool in a supported format: GnuCash XML, a Firefly III export, Beancount, QIF, OFX or a CSV file. Create accounts and secondary categories first, since every imported row must map to both. In the import dialog, review the rows highlighted as invalid, apply batch rules to map categories, accounts and tags, and import in batches by year or account. Since 2.1.0 a partial failure no longer rolls back the batch, so compare account balances against the source after each run and note which rows failed. Split transactions from GnuCash, QIF or Beancount are not carried over.

Recording daily spending from a phone

Open the instance in the phone's browser and add it to the home screen as a Progressive Web App. Use templates for recurring purchases, the Add Transaction Button widget on the mobile overview for quick entry, and scheduled transactions for rent or subscriptions. If an LLM provider is configured, pasted text such as a payment notification, or a receipt photo, can prefill a transaction, which the user still confirms. After 2.1.0, existing scheduled transactions in time zones with daylight saving time must be saved once more for the date fix to apply.

Reconciling accounts and credit cards

Set a statement date on each credit card account so the transaction list can filter the current or previous billing cycle. Open the reconciliation statement for an account, compare its closing balance with the bank statement, and then set the account's last reconciled time to the statement end date. With the editable transaction range set to the last reconciled time or later, earlier transactions become read-only. The Account Unreconciled Transactions widget added in 2.1.0 shows what still needs checking.

Analyzing spending patterns

Use the built-in statistics for category and account proportions and trends, then move to Insights Explorer for questions such as spending by hour of day, by merchant tag or by amount range. Save each exploration and export the result as CSV, TSV or Markdown for a spreadsheet or a report. AI custom chart generation sends only the prompt and the current chart code to the configured model, not transaction data, but the generated JavaScript should be read before it runs. The feature list records which release introduced each chart type and metric.

Connecting AI assistants and scripts

Enable enable_mcp or enable_api_token, generate a token for a dedicated user session and restrict the allowed IP addresses where the client has a fixed address. Point an MCP client at /mcp with the bearer token, or install the Agent Skill and its ebktools script for coding agents. Test add_transaction with dry_run before allowing writes. Remember that query results sent to a third-party model leave the server, and revoke tokens from the session list when a client is retired.