Bkper Agent
Automatically extracts transactions from uploaded documents using AI. Drag invoices, receipts, or bank statements into your book — the agent extracts dates, amounts, and descriptions, then finds the right accounts based on your transaction history.
The agent learns from your corrections and improves extraction accuracy over time.
How it works
Upload — select an account and drop a file into the book. The agent extracts transactions from the document and creates drafts. For bank statements and CSVs, it creates one transaction per line item. For invoices and receipts, it creates a single transaction.
Attach — drop a file onto an existing transaction. The agent extracts data from the document and updates the transaction with the extracted amount, date, description, and custom properties.
The agent detects the document type automatically — it first attempts statement extraction (multiple transactions), then switches to invoice extraction if the document is an invoice or receipt.
Account discovery
After extracting data, the agent assigns From and To accounts by searching your book's transaction history for similar patterns:
- Extract key terms — AI identifies meaningful parts of the description (e.g.
Office DepotfromOFFICE DEPOT #1234 PURCHASE 04/15) - Search by account + creator — finds posted transactions with similar terms, scoped to the selected account and user
- Search by account — same search, any user
- Search by description — searches across all accounts
- Pick by frequency — assigns the most commonly used accounts from matching results
When you correct an account and post the transaction, the agent re-runs discovery on all pending drafts in the book — one correction can fix many.
The agent also extracts amounts and dates from transaction descriptions using AI, even without a file upload.
Learning from corrections
When you edit an AI-extracted transaction and post it, the agent compares the posted version against its original extraction. If the data changed significantly, it improves its extraction instructions:
- Reflect — analyze what the extraction got wrong
- Curate — generate new instructions and merge with existing ones
- Validate — re-extract the document with the improved instructions and verify results are better without regressions
The improved instructions are saved to agent_prompt on the account or group. Future documents processed through that account benefit from the accumulated learning.
For invoices, this runs immediately on post. For statements, the agent flags the account and processes improvements asynchronously.
Configuration
Account and group properties
| Property | Description |
|---|---|
agent_prompt |
Custom extraction instructions for documents uploaded to this account or group. The agent appends learned instructions here automatically |
agent_prompt_id |
Identifier used to match uploaded files by filename. When a filename contains this value, the agent uses this account's prompt for extraction. Defaults to the account name |
agent_prompt_skip |
JSON array or comma-separated list of property keys to ignore during extraction comparison and learning |
agent_parser_url |
URL of an external extraction service. When set, file parsing is delegated to this endpoint instead of using built-in AI |
Example — custom extraction instructions:
agent_prompt: |-
Extract transactions from the statement table.
- Do not include header rows
- Include the reference number as a property called "ref"
Example — skip properties that vary per transaction:
agent_prompt_skip: ["trade_date", "settlement_date"]
Book properties
| Property | Description |
|---|---|
agent_parser_url |
URL of an external extraction service for all files in this book. Account and group agent_parser_url take priority |
Events handled
| Event | Behavior |
|---|---|
FILE_CREATED |
Extracts transactions from the uploaded file and creates draft transactions. Delegates to external service if agent_parser_url is set |
TRANSACTION_CREATED |
When a transaction has an attached file, extracts data and updates it. Also runs account discovery and AI-based amount/date extraction on all new transactions |
TRANSACTION_POSTED |
Compares posted data against the original AI extraction and improves the prompt if corrections were made. Re-runs discovery on pending drafts when accounts change |
TRANSACTION_UPDATED |
Same as TRANSACTION_POSTED |
TRANSACTION_DELETED |
Flags the source account for prompt improvement when a statement-extracted transaction is deleted, indicating a false positive |
BOOK_AUDITED |
Processes one flagged account or group, running the full improvement pipeline for statement extraction |
Supported file types
| Format | Extensions |
|---|---|
| CSV | .csv |
.pdf |
|
| Images | .jpg, .jpeg, .png, .bmp, .webp |
External extraction service
When agent_parser_url is set, the agent delegates file parsing to that URL instead of using built-in AI. The lookup priority is:
- Account property
- Account's group property (first match)
- Book property
The service receives a POST request with the file content (base64-encoded), book metadata, and account context, and returns structured transactions in JSON format.
This is useful for specialized parsers that handle bank-specific or proprietary statement formats.
Transaction metadata
The agent stores metadata on processed transactions. These are internal properties (suffixed with _) used for learning and traceability:
| Property | Description |
|---|---|
agent_extracted_ |
Original AI extraction data (JSON). Used to detect corrections when the transaction is posted |
agent_file_id_ |
ID of the source file. Links statement transactions to their document |
agent_credit_account_id_ |
ID of the discovered credit account |
agent_debit_account_id_ |
ID of the discovered debit account |
agent_description_part_ |
The key term that matched during account discovery |
Learn more
- Recording transactions — how to record and manage transactions in Bkper