troubleshooting-faq
FAQ
Straight answers about VaultTerm's security model, self-hosting, air-gapped operation, where keys live, supported platforms today, and how AI is kept private.
Updated Jun 23, 2026
Short, honest answers to the questions people ask most. For definitions of the terms used here, see Core concepts.
Security model
Is VaultTerm zero-knowledge?
No. VaultTerm is an audited access broker, not zero-knowledge, and we say so rather than market a guarantee we cannot keep. The server decrypts a secret in memory only for a specific authorized, audited action — revealing a value, injecting a credential into a session, scanning output for a leak — and then discards the plaintext. No plaintext is kept at rest, and every such action lands in a tamper-evident audit trail. A strict zero-knowledge design would rule out the broker features teams actually need, all of which require the plaintext in memory at the moment of use. See Security model.
Self-hosting
Can I self-host it?
Yes. Self-hosted / on-prem mode runs VaultTerm on your own infrastructure as a Docker Compose unit, including fully air-gapped. It is the same image and the same features as the hosted service; only the configuration differs. See Self-hosting overview and SaaS vs self-hosted.
Does an air-gapped install phone home?
No. An air-gapped install does not need to reach the internet. Entitlements are verified offline from a signed license file, so there is no activation call. The connected heartbeat is opt-in and off by default — an air-gapped deployment simply leaves it off. See Licensing and activation.
Where do my encryption keys live?
VaultTerm uses envelope encryption: each secret is sealed with a per-record data key, and that key is wrapped by a higher master key. Where the master key lives depends on how you configure key management — a local key, or your own HSM via HashiCorp Vault Transit. In on-prem mode the keys are yours, in your own infrastructure. See Cryptography and HashiCorp Vault sync.
Platforms
What platforms are supported today?
Today the web app is the only client you can obtain and use. The browser extension, desktop, mobile and CLI clients are built and in development, but none of them is published to a store or a public registry yet, so there is nothing for you to install. We would rather state this plainly than imply clients that are not yet shipping. See Downloads for the current per-platform status, and Browser extension overview for what the extension will do once it is published.
AI
How is AI kept private?
AI is privacy-first and LAN-default. By default, AI runs against a self-hosted Ollama model on your own network, so terminal output and secrets do not leave it. What keeps them there is that default plus a per-organization opt-in that is off until an admin turns it on — the cloud path additionally needs a configured cloud key and the cloud-AI plan entitlement, and all of them are required together.
When an organization does opt in, be clear about what travels: the commands your members ran, the terminal output around them, and the hostnames and file paths inside that text all reach the cloud model. Recognisable secrets, email addresses and IP addresses are stripped first, and every cloud call is recorded on the audit trail with a preview of exactly what was sent — but that step is a secret-and-PII filter, not a transcript filter, and it is not what keeps operational content on your network. Leaving the organization on the self-hosted default is. See AI privacy model and AI overview.