--- title: "Privacy & data" description: "What Cadence uploads, who can see it, and how contributor-private data differs from what managers see." url: "https://teamcadence.ai/docs/privacy/" --- ## Private by default Session analysis is visible to you, the developer who ran it, unless you explicitly share it. Your manager or org admin doesn't get a feed of your raw transcripts by default - they only see aggregated, org-level signal. Under the hood, every row in Cadence's database is scoped to your org, and personally-scoped data (like your own sessions) carries an additional owner-scope layer that only your account can read without an explicit share. This is enforced at the data layer, not just hidden in the UI. ## Sharing a session When you do want to share a specific session - say, to explain a tricky PR in review, or to hand evidence to a teammate - you can generate a share link or token. Anyone with that link can view that one session, without needing their own Cadence access to everything else you've done. The [CLI](/docs/cli/commands/) and the web Session Viewer both accept a shared-session token or URL directly, in addition to your own private session IDs. ## What gets uploaded Only sessions from repos you've explicitly enabled are uploaded - repo discovery during onboarding proposes enabling a repo, it doesn't do it silently. Uploads are sanitised before they leave your machine, and again server-side — see [Security](/docs/security/) for what gets redacted and how. `cadence session export` (used to build local [Cue](/docs/cue/overview/) memory or to feed other tooling) produces an author-free, server-redacted JSON export - it's designed to be safe to hand to other systems without leaking who wrote what. ## Pausing and stopping uploads Pause the daemon from the tray menu at any time - Cadence stops scanning for new sessions immediately, and nothing uploads while it's paused or while the app isn't running. Disabling a repo (or leaving it out of your selection during onboarding) has the same effect for that repo specifically. ## What Cadence Cue keeps locally [Cadence Cue](/docs/cue/overview/) builds a local memory artifact - embeddings and a graph derived from your own session history - that stays on your machine after the initial build. It's used to give your coding agent a "starting guide" before a task; it isn't a separate upload path.
Docs/Privacy & data
Privacy & data
What Cadence uploads, who can see it, and how contributor-private data differs from what managers see.
Private by default
Session analysis is visible to you, the developer who ran it, unless you explicitly share it. Your manager or org admin doesn’t get a feed of your raw transcripts by default - they only see aggregated, org-level signal.
Under the hood, every row in Cadence’s database is scoped to your org, and personally-scoped data (like your own sessions) carries an additional owner-scope layer that only your account can read without an explicit share. This is enforced at the data layer, not just hidden in the UI.
Sharing a session
When you do want to share a specific session - say, to explain a tricky PR in review, or to hand evidence to a teammate - you can generate a share link or token. Anyone with that link can view that one session, without needing their own Cadence access to everything else you’ve done. The CLI and the web Session Viewer both accept a shared-session token or URL directly, in addition to your own private session IDs.
What gets uploaded
Only sessions from repos you’ve explicitly enabled are uploaded - repo discovery during onboarding proposes enabling a repo, it doesn’t do it silently. Uploads are sanitised before they leave your machine, and again server-side — see Security for what gets redacted and how.
cadence session export (used to build local Cue memory or to feed other tooling) produces an author-free, server-redacted JSON export - it’s designed to be safe to hand to other systems without leaking who wrote what.
Pausing and stopping uploads
Pause the daemon from the tray menu at any time - Cadence stops scanning for new sessions immediately, and nothing uploads while it’s paused or while the app isn’t running. Disabling a repo (or leaving it out of your selection during onboarding) has the same effect for that repo specifically.
What Cadence Cue keeps locally
Cadence Cue builds a local memory artifact - embeddings and a graph derived from your own session history - that stays on your machine after the initial build. It’s used to give your coding agent a “starting guide” before a task; it isn’t a separate upload path.