Security, privacy and data
How your data is handled.
Mparanza is designed to stay out of the path of your work. The important question is not whether you trust a promise, but which systems actually receive your data.
The local data boundary.
Vera and Clara add specialist methods to the Codex environment you already use. They do not add a Mparanza application server between your workspace and OpenAI.
Mparanza does not receive the working content.
The scripts run on your machine.
Vera and Clara's tools execute from the Codex workspace on your computer, using the local files, software, and permissions you control.
This lets you analyze, filter, and aggregate data locally, and limit what you send to the model to the information needed for the request.
Unlike a conventional ChatGPT data-analysis workflow, where uploaded files are made available to a provider-managed Jupyter notebook, these scripts run where the files already live.
One fewer system to trust.
Model requests use your existing Codex/OpenAI account. Mparanza is not the intermediary and cannot inspect prompts, files, or outputs it does not receive.
The terms, data controls, and workspace policies attached to your account continue to apply.
GDPR follows the actual data flow.
A local-first architecture can reduce the number of recipients and copies, but it does not by itself answer every GDPR question. Local workflows do not add Mparanza as another recipient of your working content.
Where Mparanza processes personal data, our policy explains the scope, purpose, retention, and rights that apply.
Hosted features are explicit.
If you choose an optional Mparanza-hosted feature, the content needed to provide it reaches systems controlled by Mparanza. We identify that boundary and explain the applicable retention and deletion rules in our policy.
Services you connect yourself remain governed by their own permissions and terms.
Verify the position.
You do not have to rely on the claim alone.
Secure by design. Trust minimized, not merely promised.