Skip to content

These pages describe PrivacyFence 7.0.2. Download it · What changed

PrivacyFence documentation

How to install, use, deploy and secure PrivacyFence. These pages describe PrivacyFence 7.0.2.

Install and first steps

Using PrivacyFence

  • How PrivacyFence works — the daemon, the companion app, how AI clients connect, and the privacyfence_* tools every client sees.
  • Approvals and policy — how requests are gated, approval cards, the PII check, "Always allow" and policy rules, the privacy filter, file previews and notifications.
  • Configuration reference — every settings.yaml key and every organization-bundle option, with its default.
  • privacyfence-sync — run a scheduled data sync as a recipe, with no AI model involved: install, connecting, the rules it needs, and the recipe-writing plugin.

Organization deployment and security

  • Organization deployment — deploy PrivacyFence centrally for an organization, including claude.ai as a client.
  • Organization deployment with Docker — run the organization deployment as containers: the published image, a tested Compose file with Caddy, upgrades, backups and verifying the image.
  • Sign in to other apps with PrivacyFence — let Peek and Grist (or other apps you name) sign people in through PrivacyFence, with the identity PrivacyFence already has.
  • Security and compliance — trust boundaries, step-up authentication, privilege separation, audit integrity, and what PrivacyFence does not claim.
  • OAuth relay runbook — the OAuth redirect relay at oauth.privacyfence.eu: what it forwards and stores, and how the maintainers run, check and revoke it.

Connector setup

Registering your organization's own OAuth app (or bot) for a service, in place of the built-in one, as a signed connector config on an organization server or a bundle section on a desktop install:

AI agent setup

Connecting each AI client, in local mode and in an organization deployment:

Reference appendices

Release history is in ../CHANGELOG.md.