Camelot and SpendixAI both start with a bank statement PDF, but they end in different places. Camelot is usually described as an open-source Python library for extracting PDF tables, run locally. SpendixAI is an expense tracker that imports statements and turns them into categorised spending. One is a building block, the other is a finished tool.
Who is each one for?
| Point | Camelot (usually described as) | The app |
|---|---|---|
| Type | Python library, open source | Expense tracker app |
| Where it runs | Your own machine | Web plus mobile apps |
| Output | Tables you process yourself | Categorised transactions and reports |
| Skills needed | Comfort with Python | None; review rows on screen |
| Cost | Open source | Free plan; Pro from ₹99 a month |
The comparison is limited to what both sources support. For anything else about Camelot, check its own documentation.
What happens after the table is extracted?
With a library, extraction is step one of several. You still need to clean the rows, parse dates and amounts, decide categories and build any summaries. That is a rewarding project if you enjoy it, and flexible if you want a custom format.
In the app the same statement goes in through an upload, you review rows before saving, and categorisation runs in stages: your own rules, then your history and similar merchants, then an AI model only if AI is on. Corrections become history, so the next import needs fewer fixes.
Can I use both?
Yes. Some people extract with a script for an odd document and import a CSV into a tracker afterwards. The app accepts CSV and Excel as well as PDF, so a cleaned file from any source can be imported.
How do the privacy models differ?
A local library keeps the PDF on your computer. The app asks you to upload the statement, but it never connects to your bank, never asks for net-banking logins, does not read SMS and uses Google sign-in only. AI is optional and can be limited by feature or by a monthly budget; an optional feature that reads unfamiliar layouts does process the uploaded document.
A worked example of the time trade-off
Say you handle 12 statements a year. If a script needs about 40 minutes of tinkering to fix its output for each, that is 12 x 40 = 480 minutes, or 8 hours. If an importer needs about 10 minutes of row review each, that is 12 x 10 = 120 minutes, or 2 hours. These are hypothetical figures: your own timings will differ, so time one real statement each way.
What to verify before choosing
- Do your statements parse cleanly? Try one from each bank or card.
- Do the extracted totals match the document?
- How many fixes does each month need, and does the tool learn from them?
- Are you comfortable with where the file is processed?
What does reviewing imported rows involve?
Whatever route you take, the same few problems show up in bank statement tables. Knowing them lets you spot errors quickly instead of reading every row.
| Problem | What to look for | Quick fix |
|---|---|---|
| Wrapped descriptions | One transaction split over two lines | Merge the lines before importing or fix the row on review |
| Debit and credit columns | Amounts landing in the wrong direction | Check a known payment and a known deposit |
| Date formats | Day and month swapped | Look at a date such as the 13th, which cannot be a month |
| Opening balance rows | A balance line counted as a transaction | Delete the line |
| Repeated headers | Header text on each page treated as data | Remove those rows |
Ten minutes of checking on a first import saves a month of wrong reports later.
A sensible decision rule
If you enjoy code and want control, use the library. If you want to open a dashboard after uploading and see where the month went, use an importer. Neither choice is permanent.
Useful ideas start with a clearer picture of your own spending.
Explore your reports


