-
Notifications
You must be signed in to change notification settings - Fork 2
PRIVACY POLICY
Effective date: 2026-09-14
Product: Graphene Agent Bridge (Store product id 9P61PN50Q957)
Publisher: Graphene-Lab
We do not collect your data. Not your name, not your files, not your prompts, not your settings. The product has no account, no sign-in, no registration, and no telemetry. Nothing is sent to us, ever.
The program runs on your own machine. Your data stays there.
Nothing.
- No analytics, no advertising, no usage tracking.
- No phone-home and no update channel of our own — updates come from the Microsoft Store.
- The Package Support Framework inside the package was compiled from Microsoft's public source code on purpose, with its telemetry provider identifier left as the empty placeholder. The binaries from the NuGet package, which carry Microsoft's real telemetry destination, were rejected. So there is no address to send telemetry to, even if a component tried.
Everything stays on your machine:
| Location | What it holds |
|---|---|
PersistentData\ |
your settings, including model provider names and the API keys you typed |
attachments\ |
files you attached to a conversation |
tui-screenshots\ |
screenshots you took from the terminal |
GiraffeAIWebClient\ |
the bundled web client's files |
In the Store package these folders are redirected into your own user profile
(%LOCALAPPDATA%\Packages\41836WindowsPhne.GrapheneAgentBridge_6mfwch41bwvk2\LocalCache\),
because the package itself is read-only. They are not copied anywhere else.
Please treat the API keys in PersistentData\ as secrets and keep that machine protected.
AgentBridge was made for people and organisations that do not want to depend on outside AI services. You run it yourself. You choose what it talks to.
For the strongest privacy, run a local model on your own hardware. In that setup your prompts and your files never leave your machine at any point. This is the intended way to use the product when privacy is the goal. You need enough local hardware for the model you want to run.
The product can also talk to an external model provider. When you do that, your content goes from your machine to the provider you picked, using the API key you supplied. We are not in that path and we cannot see it.
That provider's privacy policy governs your data from that moment. Before connecting one, read its privacy policy and decide whether you accept it. The choice, and the key, are yours.
AgentBridge has a built-in anonymisation feature for this case
(LLM:Anonymize, on the command line: --LLM:Anonymize true).
When it is on, the program replaces personal details — names, keys and other sensitive identifiers — before the request is sent to the outside model, and puts them back when the answer comes. The agent keeps working at full strength, but the outside machine never sees what should stay private. It adds no noticeable delay and is invisible in normal use.
This is optional. It is there for people who use an outside provider and still want to keep personal details out of the request.
Note: we do not collect the anonymised data either. Anonymisation protects what you send to the provider you chose. It is not a collection channel, and nothing arrives at us in any form.
This is a professional office tool. It is not made for children.
If this policy changes, the new version will appear at the same address with a new date.
Graphene-Lab — https://graphenelab.it