The Bkper Agent automatically extracts and categorizes Transactions from uploaded documents using AI, eliminating manual data entry. Simply upload invoices, receipts, bank statements or csv files, and the Agent intelligently parses amounts, dates, descriptions, and finds the right accounts in your book.

The Agent uses artificial intelligence to:

- Extract transaction data from documents (PDFs, images, CSV files)

- Automatically detect whether you uploaded an invoice, receipt, or bank statement

- Learn from your existing transactions to match your bookkeeping patterns

- Discover and assign accounts by analyzing similar transactions in your book

- Self-improve over time by learning from your edits on Transactions it recorded

The Agent processes three types of documents:

Bank Statements & Credit Card Statements

- Extracts multiple transactions from a single statement

- Supports CSV files and scanned/photographed statements

- Automatically organizes transactions by date

Invoices

- Extracts single transaction with vendor/customer details

- Identifies invoice numbers, dates, and amounts

- Captures custom fields like tax amounts or invoice references

Receipts

- Extracts purchase details and amounts

- Identifies merchant names and transaction dates

- Perfect for expense tracking

Make sure you're using the alpha interface to access this feature.

The installation takes less than a minute:

Sign into your Bkper Account and open the book where you want to use the Agent. Click the Settings button (⚙️) in your book's navigation, then select Automations from the menu. You'll see the Bkper Agent listed among available apps. Click Install, and you're done—the Agent is now active and ready to process your first document.

Now you can simply drag and drop a document to start experimenting with the intelligent document parsing that learns from your edits.

The simplest way to use the Agent is to upload documents directly:

Select the Account related to the file you are uploading. for example Select the Brex Cash Account in your book when you upload a statement from your Brex account.

Drag and drop a file (PDF, image, or CSV) into the transaction area of your book.

The Agent automatically processes the file and draft Transactions are recorded and the other Account is completed based on the Transaction history in this book.

Review and Post the Transactions.

Note: For bank statements and CSV files the Agent creates multiple draft Transactions (one per line item). For invoices, it creates a single draft Transaction.

You can also attach documents to existing transactions to enrich them:

Drag and drop a file directly onto that transaction, the Agent extracts data from the file and updates the Transaction with the extracted details.

This method is perfect for adding invoices to existing expense entries or attaching receipts for documentation.

The Bkper Agent gets smarter over time through a process called Agentic Context Engineering (ACE):

What the Agent Learns From:

1. Your existing transactions - The Agent analyzes patterns in your transaction history

2. Your corrections - When you edit AI-extracted data, the Agent learns what's correct

3. Your account structure - It understands how you organize your chart of accounts

How Corrections Work:

1. Edit the transaction as needed (change amount, date, description, or accounts)

2. Post the corrected transaction

3. The Agent automatically flags the extraction for improvement

4. Behind the scenes, the Agent analyzes what went wrong

5. It updates its extraction instructions to avoid the same mistake

6. Future documents are processed with improved accuracy

This learning process happens automatically—no configuration required!

One of the Agent's most powerful features is Account Discovery. When extracting transactions, the Agent automatically suggests From and To Accounts by:

- Analyzing transaction descriptions and extracting key terms

- Searching your transaction history for similar patterns

- Matching based on your past behavior (e.g., "Uber" always goes To Transport Expense)

- Learning from the creator (your transactions vs. your team members' patterns)

The more you use Bkper, the better the Agent becomes at finding the right accounts!

While the Agent works great out-of-the-box, you can customize its behavior for specific use cases:

Custom Extraction Instructions

Add an agent_prompt property to Accounts or Groups to control how the Agent extracts data:

Example: You have a "Bank Statement - Chase" account and want to extract specific details:

1. Open the Account settings

2. Add a custom property:

- Key: agent_prompt

- Value:

Extract transactions from the statement table.

- Include the reference number as a custom property called "ref"

- Extract merchant category if available

3. Save the Account

Now, when you upload a Chase bank statement with this Account selected, the Agent follows your custom instructions!

For teams managing multiple books, you can store extraction instructions in a central "prompt book":

1. Create a dedicated book for storing prompts (e.g., "Company Extraction Rules")

2. In your operational book, add a book property:

- Key: agent_prompt_book_id

- Value: ID of your prompt book

This allows you to maintain consistent extraction rules across all your books.

Account-Specific Prompts

Match prompts from the remote book using agent_prompt_id:

- Key: agent_prompt_id

- Value: Name or ID to match in the remote prompt book

This is useful when you have different types of statements (bank, credit card, etc.) that need different extraction rules.

For Bank Statements:

- Upload statements regularly (monthly or weekly)

- Select the bank account before uploading for better account discovery

- Review and post draft transactions in batch

- Correct any errors—the Agent will learn from them

For Invoices and Receipts:

- Select the related account before uploading for better account discovery

- Upload immediately after receiving

- The Agent works best with clear, readable scans

- Include the account (e.g., "Office Supplies") in the upload context

- Add custom properties if you need specific fields extracted

For Maximum Accuracy:

- Post transactions regularly so the Agent has more data to learn from

- Be consistent with account names and transaction descriptions

- Correct mistakes immediately to trigger learning

- Use groups to organize accounts by document type (e.g., "Bank Accounts", "Credit Cards")

Here's how a typical workflow looks with the Bkper Agent:

D day: Upload Your First Bank Statement

1. Select the Bank Account in your book

2. Drag your bank statement or CSV into the Transaction area

3. The Agent records as many Draft Transactions as line items in the statement

4. Review the Transactions—most accounts are likely missing (Agent is learning)

5. Assign accounts manually and post Transactions

D+ : Upload Second Statement

1. Select the Bank Account in your book

2. Upload the next statement

3. The Agent now auto-assigns accounts for ~60% of Draft Transactions

4. Review and correct the remaining 40%

5. Post all transactions

D++ : Upload Third Statement

1. Select the Bank Account in your book

2. Upload the statement

3. The Agent auto-assigns accounts for ~85% of transactions

4. Only minor corrections needed

5. Post in minutes instead of hours

D+++: Fully Automated

1. Select the Bank Account in your book

2. Upload statement

3. Agent correctly extracts 95%+ of transactions

4. Quick review and post

5. Your bookkeeping is done in an instance!

The Agent stores metadata on each transaction to help you track its learning:

- agent_extracted_ - Original AI extraction (before corrections)

- agent_credit_account_id_ / agent_debit_account_id_ - Discovered accounts

- agent_description_part_ - Key terms used for discovery

- agent_file_id_ - Links transaction to source document

You can view these properties in the transaction details to understand how the Agent processed each entry.

Agent didn't extract anything

- Verify the Agent is installed (Alpha → Settings → Automations)

- Check file format (must be PDF, image, or CSV)

- Ensure file is readable (clear scan, not corrupted)

Extracted data is incorrect

- Simply correct the Transaction and post it

- The Agent will learn from your correction

- Future extractions will be more accurate

Accounts not auto-assigned

- This is normal for new books with limited Transaction history

- Keep posting Transactions—the Agent needs data to learn

- Try selecting an account before uploading files

Same mistakes keep happening

- Make sure you're posting corrected Transactions

- Check if you're editing and saving (not just viewing)

- The Agent learns incrementally—give it a few correction cycles

Privacy & Security

- All file processing uses Google Gemini AI with enterprise-grade security

- Files are processed through Cloudflare's AI Gateway for additional protection

- Your data never trains public AI models—it only improves your private Agent

- Files are stored securely in your Bkper book

- You can delete files at any time

- The Agent learns per book so each learn path is unique

The Bkper Agent transforms document processing from hours of manual data entry into minutes of review time. By combining AI extraction with continuous learning (ACE), it adapts to your unique bookkeeping style and gets smarter with every transaction you post.

Parse Invoice

Parse Bank Statement