Visual FoxPro in 2026: your four exit options, honestly compared
Visual FoxPro has been out of all Microsoft support since 13 January 2015, and Microsoft confirmed in 2007 that no version 10 would ever come. A VFP system in 2026 has four realistic options: stay and contain the risk, move the data to a modern database and keep the VFP front end, replace with an off-the-shelf package, or rebuild as a web application behind a parallel run. The right one depends on how bespoke your system is and how close your tables are to the 2 GB ceiling.
Visual FoxPro is the politest of the dead platforms. It rarely crashes, the DBF data layer is open and well documented, and systems built on it are still quietly running businesses. That politeness is exactly why so many owners have not moved: nothing has forced the question yet.
Here is the question anyway, with the four honest answers. The dates below are Microsoft’s own; the full set, with sources, lives on the Visual FoxPro migration page.
Where VFP actually stands in 2026
- Microsoft announced in March 2007 that there would be no VFP 10. Version 9, released in December 2004, is the end of the line.
- Mainstream support ended on 12 January 2010. Extended support ended on 13 January 2015. Nothing has been patched since.
- No Windows release after Vista was ever officially supported for VFP 9. Installations on Windows 10 and 11 run outside any support arrangement.
- The product is 32-bit forever, with a hard 2 GB per-table limit Microsoft declined to lift.
- IT Jobs Watch recorded zero permanent UK job adverts citing FoxPro in the six months to June 2026.
None of that means your system stops working tomorrow. It means every year of staying is a year of narrowing options, and the narrowing is not linear: it ends abruptly, usually with a Windows update or a resignation letter.
The four options, compared
| Option | What it means | What it costs you | Honest verdict |
|---|---|---|---|
| Stay | Keep running VFP 9 on unsupported Windows, contain the risk | Rising fragility, Cyber Essentials exposure, an eventual forced exit on someone else’s timetable | Defensible only with tested backups, an imaged machine and a written exit date |
| Refactor the data layer | Move tables to SQL Server or similar, keep the VFP front end | Real engineering work that still leaves an unsupported, unstaffable front end | A genuine stopgap, worth it when tables are near 2 GB and the rebuild needs phasing |
| Replace with a package | Buy an off-the-shelf system, migrate the data into it | Decades of bespoke logic forced into someone else’s shapes, per-user fees forever | Right for systems that were never very bespoke to begin with |
| Rebuild as a web application | Extract the data, rebuild the behaviour, run old and new in parallel, cut over | The largest one-off cost, then ownership: no licences, no ceiling, any developer can maintain it | The exit. The parallel run is what makes it safe |
Two notes on that table that vendors of each option will not volunteer.
First, the refactor row is undersold by migration firms and oversold by FoxPro specialists. Moving the data to a proper server genuinely fixes the scariest failure modes, corruption on shared DBF files and the 2 GB wall, and it is honest work. But it leaves you hiring from the same empty talent pool for the front end, so it postpones the people problem, which is the one that actually kills these systems.
Second, the package row depends entirely on a question owners rarely ask: how bespoke is this system really? Twenty years of small customisations feel unique from inside the business. If the audit shows the system is 80 per cent a standard sales ledger, a package plus a small integration is the cheap, right answer. We will say so when it is; it is less work for us and a better outcome for you.
How to decide in one afternoon
Answer three questions, honestly, with the people who use the system daily.
- How close is the biggest table to 2 GB? Check the file sizes. Over about 1.5 GB, you are choosing under time pressure whether you admit it or not.
- Who can change the system today, and what is their retirement date? That person’s calendar is your real support contract.
- Which screens would a package have to replicate exactly? If the answer is “most of them”, the package option is an expensive way to discover you needed a rebuild.
If the answers worry you, that is the moment for a free Legacy Risk Audit: a 30 minute call and a one-page written risk summary, including which of the four options the evidence points at. No obligation, and the one-pager is useful board ammunition either way.