A Python tool for replicating Microsoft Access databases to PostgreSQL, MS SQL Server, MySQL or MariaDB
This project has the objective of replicating any MS Access database into the equivalent PostgreSQL database, MS SQL Seerver, MySQL or MariaDB with as little configuration as possible.
In the AI research world, there is almost no development aimed at integrating with user data stored in MS Access databases. There is however significant support for PostgreSQL database and others, so by creating a replicate of the user data, with the option to update that replica at will, it becomes possible to connect AI platforms to the data normally being managed using MS Access.
This document describes the ‘replicator’ program – not the research tools and processes that might use the replicated data.
There are three versions of the replicator.
- pg_replicator.pg - the target database is in a PostgrSQL server
- ms_replicator.py - the target database is in an MS SQL server
- my_replicator.py - the target database is in a MySQL server (or MariaDB)
MS Access is a very effective UI and data management tool, but has strong limitations related to capacity and robustness. Microsoft recommends MS SQL Server for users where those limits become problem. A consequence of that is that as of 2026, Microsoft has marked a number of re-distributable support libraries for Access database as ‘no longer supported’. That does not mean MS Access is unsupported, but given that the UI component of Access can use MS SQL Server as its database, and MS SQL Server has an ‘Express’ which has higher capacity limits than MS Access and is free for use, you can deduce that the MS Access database component is being strongly de-emphasized.
I would estimate that many many users of the original MS Access tool have only modest transaction and storage capacity requirements, and very limited budgets for software migrations or experiements, so a free tool like this allows them to explore the amazing amount of development in the AI community for analysing data.
This replicator is a bare bones equivalent of MS SQL Server Migration Assistant but for a target PostgreSQ, MS SQL database,MySQL or MariaDB, which is then accessible from many AI integration platforms.
However the intended use is not the same as that Migraton Assistant
The replicator is just a normal program. But its functions allow you to easily maintain a replica of your data in a database system that is much better (directly) supported in a wide range of AI platforms.
We did experiment with Microsoft SQL Server Migration Assistant, but that is aimed at migrating away from MS Access rather than maintaining a replica - and it had lots of problems with some elements of the Access system we had as input – the same problems we had. Primarily with stored queries It did not convert any of them, and we do not replicate them either. Microsoft SQL Server Migration Assistant is an interactive migration assistance tool rather than a production process.
A big point of this approach is to leave your existing Access application(s) untouched. As a 'single source of truth' ... and the replication process becomes an automated workflow with zero manual steps to avoid that source of errors.
Lots of people are extracting Excel or CSV files and feeding them into an AI platform. That works of course, but is normally a manual approach - fine for experiments, but not so good for regular operation.
Refer to the installation notes in the relevant PDF file in the /docs directory
Although this is an MIT licensed project, the documentation is explicitly excluded from that license.
It stands alone as copyright documents delivered in .pdf format and markdown format. It is available for you to read under the usual fair use arrangement for this sort of documentation.
We have done this to encourage the use of this repository as a central point to coordinate revisions ... and retaining control of the docummentation encourages people to work on this repository instead of just forking their own version of the code. But of course people remain free to do that too.