Data and Privacy
Project files
Genexy works with folders on your device. Editing, local Git operations, terminals and Docker all act on that local workspace. Nothing requires uploading the project. The Web IDE is not hosted either: the browser connects straight to your own Genexy IDE, which serves the interface.
What leaves your device
| Action | Destination | What is sent |
|---|---|---|
| AI chat | The agent or provider you selected | Your prompt, attached context and tool results |
| Vulnerability scan (on request) | OSV.dev | Dependency names and versions |
| Language server installation | The public registry or GitHub release of that server | The package request only |
| OpenAPI / AsyncAPI validation | spec.openapis.org, asyncapi.com | A request for the official JSON Schema |
| Remote access, direct | The IP address and port you entered | The encrypted session between your two devices |
| Remote access over WebRTC | Genexy signaling service, a public STUN server | Connection setup: device identifiers, session descriptions and network addresses |
| Remote access, relayed | Genexy services (relay only) | The encrypted session data, forwarded between your devices |
| Support ticket | Genexy services | Your description, comments and attachments |
| Sign-in, downloads, profile | Genexy services | Account data required for authentication and entitlements |
Everything else (file operations, search, Git against a local repository, terminals, Docker) stays on the device.
Genexy does not proxy AI traffic. The IDE talks directly to the agent or OpenAI-compatible endpoint you selected, which means that provider’s terms, retention and training policy are the ones that apply to your prompts. Review them before selecting a provider; a free provider is still a third party.
Remote access
Remote Access connects two of your own devices in one of two ways:
- Directly to an IP address and port. The client connects straight to the host. No Genexy service takes part.
- Over WebRTC. Genexy services only help the two devices find each other (signaling), and a public STUN server tells each device its public network address. The devices then try to open a peer-to-peer connection, so the session data goes straight from one device to the other.
If a peer-to-peer connection cannot be established, for example behind a strict NAT or firewall, the session falls back to a relay through Genexy services. In that case Genexy acts only as a proxy that forwards the data between your devices: WebRTC encrypts the session between the two devices, and the relay neither stores nor processes its content.
If no traffic may leave your network, use the direct connection over a local network or a VPN.
AI chat storage
Conversations and attachments are stored locally. On desktop:
~/.genexy/chatsMobile operating systems keep the equivalent data inside the application sandbox.
Approvals and Autopilot
Manual approval decides whether a requested action runs. Autopilot answers those prompts for you. Neither changes the network destination of a request or the provider’s data policy; they only change who confirms an action.
Account and support data
The portal stores the account data needed for authentication, email verification, your plan and profile management. Anything you submit to Support is sent to Genexy services so the team can process the ticket.
Good practice
- keep secrets out of source files and chat messages; use your platform’s credential tooling;
- review Git diffs after AI-assisted edits before committing;
- use manual approval instead of Autopilot in sensitive projects;
- remove credentials from tickets and Discord messages;
- in environments where nothing may leave the network, do not run vulnerability scans or install language servers from the IDE; both send requests to the internet (see the table above).
Git ignore rules stop a file from being committed. They do not stop an explicitly opened file, a pasted value or a file an agent chooses to read from reaching an AI provider.
For the legal policy, see the Genexy Privacy page .