Automating software without an API: six ways in

Your invoicing, stock or point-of-sale software connects to nothing, and somebody copies its data by hand. Automating software that has no API is still possible as long as it lets its data out in one way or another. The real question is where to take the data from, and what keeping that route working will cost.

This guide ranks the six doors, from the most solid to the most fragile. Choosing the tool and going into production are covered in the other guides in our Automation and bots section, and our guide Automation without AI: when a rule is enough helps you tell whether a step needs AI or a simple rule.

First of all, one check that saves weeks of tinkering: does the software really have no API, or does it reserve it for a higher plan? On Odoo’s pricing page, viewed on 4 October 2026, access to the external API is part of the “Custom” plan, not the “Standard” plan. Comparing the price of the higher plan with the cost of a fragile workaround sometimes gives an obvious answer.

Door 1: read a scheduled export

Many business applications can export a file (CSV, Excel, XML) at a fixed time, or drop it in a shared folder or on an FTP server. It is the most solid door after an API: an export format changes less often than a screen, and a change of format is spotted as soon as the file is read.

The automation watches the folder, reads the file, compares it with the previous day’s version and processes only what has changed. In n8n, the reading is done with the Extract From File node, which turns a CSV, a spreadsheet or a PDF into usable data.

Honest caveat: the information is only as fresh as the export frequency. If the export runs once a night, your CRM is always a day behind: perfect for a morning summary, not enough for a real-time alert.

Door 2: read the emails it sends

A piece of software that talks to no other often sends emails: order confirmation, invoice, notification of a new case. These messages contain structured data, which you only need to read.

Create a dedicated mailbox and have these notifications delivered to it: a trigger reads each message as it arrives. According to the n8n documentation, that is the job of its Email Trigger (IMAP) node. The body of the message or the attachment is then parsed to extract the useful fields.

Honest caveat: if the vendor changes the email template, the parsing breaks without warning. You therefore need a control rule (a missing required field triggers an alert) rather than blind trust.

Door 3: read its database

When the software is installed on your own premises, its data lives in a database (SQL Server, MySQL, PostgreSQL). Reading it directly gives fresh, complete information, without depending on the interface.

Two absolute rules: read-only, with an account that has no other rights, and never write to the database of third-party software.

Honest caveat: a database structure is not a contract. An update can rename a table. And some vendors forbid direct access in their terms of use or support terms: check before you start, not after.

Door 4: extract the data from its documents

Invoices, delivery notes, statements: when the only output is a PDF, you read the PDF. A PDF generated by software, not scanned, contains text that can be extracted cleanly. A scanned document needs character recognition, sometimes supplemented by an AI model to find the fields.

This is the door for entering supplier invoices:

Honest caveat: extraction is never one hundred per cent right. A serious automation provides a human review queue for doubtful cases, and measures how many documents go through it: if that share does not fall over time, the gain is illusory.

Door 5: drive its interface in a browser

When nothing else exists, a program can do what the person does: open the page, log in, click, fill in, download. Tools such as Playwright drive a real browser; RPA tools, short for robotic process automation, apply the same idea to desktop applications.

It works, but it is the most fragile door, and you need to know that before you start:

• a button moved, a label changed, a consent window added, and the script fails;
• two-factor authentication or a captcha can block it completely;
• the terms of use of some services forbid automated access.

Honest caveat: a browser script needs active monitoring and a budget for fixes. If the vendor updates its interface every month, the maintenance cost can exceed the gain.

Door 6: prepare the re-entry

Sometimes the right answer is not to automate entry into the software, but everything around it. The data is prepared, checked and put in the order of the screen, and the person doing the final entry has nothing left to look up or recalculate. Less spectacular, but robust, and with no risk towards the vendor.

How do you choose the right door?

To decide, ask yourself these questions. How fresh does the data need to be, to the hour or to the day? Who will be told if it breaks, and how quickly? How many times a year does the vendor change its interface or its formats?

Whichever door you use, an automation that goes by a roundabout route must be monitored: an alert on failure, a history of runs, the ability to replay. The set-up is detailed in our guide to monitoring your n8n workflows. And if you are hesitating over the tool that will orchestrate all this, our comparison of n8n, Make and Zapier for an SME sets out the criteria.

Where to start

Start with the task, not the tool: note what the person copies, where the data comes from and how often. The right door can often be read off that list.

If your business software talks to nothing, that is exactly the case for our offer to remove manual re-entry with n8n: we first observe the task with the person who does it today, we build the automation and test it on your data, then it takes over gradually, with the old way kept as a fallback.

Frequently asked questions

Sources

1. Odoo’s pricing page (new tab)odoo.com
2. Extract From File (new tab)docs.n8n.io
3. Playwright (new tab)playwright.dev