A Microsoft 365 toolkit with client-side encrypted credentials and vault — Argon2id, XSalsa20-Poly1305, X25519, Ed25519 — and a list of what the server sees.
Frequently asked questions
What does "zero-knowledge" mean for GTools.pro?
The encryption key is derived from your passphrase via Argon2id and never leaves your machine. Vault contents and OAuth client secrets are encrypted with this key, and the server stores only their ciphertext. That covers storage, not use: On Cloud, when you run a tool, your browser hands the decrypted credential or token to the service for that session, where it is held in memory. Migrations run on that in-memory session. Scheduled reports, where enabled, rely on a token the server can use. Cloud exports are readable by the service until your plan’s retention window ends.
Which cryptographic primitives are used?
Argon2id (key derivation), XSalsa20-Poly1305 (symmetric encryption for credentials and vault data; XChaCha20-Poly1305 for encrypted export bundles), X25519 (key exchange), Ed25519 (signatures).
What does the server still see in plaintext?
OAuth handshake parameters and tokens until your browser collects them, credentials and tokens in use during a Cloud session, cloud exports during retention, license JWTs, migration job settings including user mappings, scheduled-report orchestration metadata, billing data and operational telemetry. Scheduled reports, where enabled, rely on a token the server can use. Documented at gtools.pro/security-overview.
Where can I report a security vulnerability?
security@gtools.pro. See gtools.pro/trust for the responsible-disclosure policy.
Threat-model limitations — what zero-knowledge is not for
Zero-knowledge here covers vault contents and credentials at rest: the GTools.pro server stores them as ciphertext it cannot open. It does not cover credentials in use on Cloud, which your browser hands to the service for the session and which are held in memory. It does not cover migrations, which run on that in-memory session, or cloud exports, which are readable by the service until your plan’s retention window ends. Scheduled reports, where enabled, rely on a token the server can use. It also does not mean immunity to operator-side compromise: a malicious browser extension, a keylogger on the operator workstation, or a stolen recovery passphrase still grants full access to the data the operator can see. The desktop edition raises the bar (vault sits inside the OS keychain / Stronghold), but the operator endpoint remains in scope for the threat model.
Server-side metadata that GTools.pro must process — billing identifiers, schedule timestamps, audit-log entries — is documented in docs/SECURITY.md. We log enough to invoice, schedule, and debug; we do not log tenant contents. Anything outside that boundary is intentionally out of scope.