skip to content

With Zero Data Retention enabled for Claude Code on Claude for Enterprise, what is and isn't covered, and how do you keep developers' sessions inside it?

level: seniorimportance: should knowfreq 6%

answer

  1. inference only, one organization at a time
  2. the account team enables it
  3. storage-dependent features switch off
  4. personal logins and other orgs escape
  5. forceLoginMethod with forceLoginOrgUUID

basics

~20 s

ZDR covers Claude Code inference for one enabled Enterprise organization: prompts and responses aren't stored after the response. Claude.ai chat, Cowork, MCP servers and admin data aren't covered; pin logins to that organization with managed forceLoginMethod and forceLoginOrgUUID.

solid answer

~40 s

Zero Data Retention for Claude Code is enabled per organization by Anthropic's account team for qualified Claude for Enterprise accounts — it isn't in the standard plan or the admin settings. It covers **inference**: prompts and responses are processed and not stored after the response returns, except where law or misuse requires, and flagged sessions may be kept up to 2 years. It does not cover claude.ai chat, Cowork, analytics metadata, seat management or data that MCP servers and other integrations receive, and features that need stored transcripts — cloud sessions, Remote Control, `/feedback` — are switched off. Coverage follows the login: a personal account or another organization's API key escapes it, so deploy `forceLoginMethod: "claudeai"` and `forceLoginOrgUUID` through device-managed settings. Local transcripts under `~/.claude/projects/` are unaffected.

code

json · 5 lines
json
{
  "forceLoginMethod": "claudeai",
  "forceLoginOrgUUID": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
  "cleanupPeriodDays": 7
}

go deeper

for a junior

Recall that zero data retention is an Enterprise arrangement in which Anthropic does not store your Claude Code prompts and responses after answering.

for a middle

List what ZDR covers and excludes, and which features stop working because they need stored transcripts.

for a senior

Show how coverage depends on the login, and deploy forceLoginMethod and forceLoginOrgUUID through device-managed settings to keep sessions in the organization.

for a principal

Treat ZDR as one control in a retention policy: integrations, local transcripts and cloud-provider terms still need their own decisions.

## What ZDR is and how it is switched on **Zero Data Retention (ZDR)** for Claude Code is an arrangement for qualified **Claude for Enterprise** organizations. With it, the prompts sent and the responses Claude generates during Claude Code sessions are processed in real time and **not stored by Anthropic after the response is returned**, except where needed to comply with law or combat misuse. If a session is flagged for a Usage Policy violation, the associated inputs and outputs may be retained for up to 2 years. It is not a toggle an administrator finds: - ZDR is **not part of the standard Enterprise plan** and can't be enabled from admin settings; Anthropic's account team enables it after confirming eligibility, and every enablement is audit-logged. - It is **per organization**: a new organization under the same account needs its own enablement. - It applies only to **Anthropic's direct platform**. Deployments through Amazon Bedrock, Google Cloud's Agent Platform or Microsoft Foundry follow those platforms' retention policies. ## Covered versus not covered | Area | Under ZDR? | | :- | :- | | Claude Code model inference in the enabled organization | Covered | | Chat on claude.ai | Not covered — standard retention | | Cowork sessions | Not covered | | Claude Code analytics | Not covered — no prompts stored, but productivity metadata such as account emails is kept | | User and seat management data | Not covered | | MCP servers and other third-party integrations | Not covered — review each service's own data handling | ## Features that switch off Anything that needs Anthropic to store prompts or completions is disabled at the backend, whatever the client shows: - **Cloud sessions**, including ones started from the desktop app - **Remote Control**, which stores the transcript to sync devices - **Feedback submission**: `/feedback`, `/bug` and `/share` - **Artifacts** and Claude Tag Trying one returns an error saying the organization's policies don't allow the action. Some models that require data retention by default may be absent or disabled in the `/model` picker, and the server rejects requests for them regardless of client configuration. ## Keeping sessions inside the ZDR organization ZDR applies to requests that **authenticate into the ZDR-enabled organization**. A developer signed in with a personal account, or using an API key from a different organization, is outside it. The managed login keys close that: 1. Set `forceLoginMethod` to `"claudeai"` so only claude.ai accounts can log in on first-party login paths. 2. Set `forceLoginOrgUUID` to the ZDR organization's ID (one UUID, or an array of several). Only a managed source enforces it; elsewhere a single UUID merely pre-selects the organization. 3. Deliver both through **device-managed settings** — MDM or the system `managed-settings.json`. Server-managed settings reach only accounts already authenticated into the organization, so they can't redirect a first login. If you also use server-managed settings, put the keys in both places, because they are not among the keys Claude Code reads from every managed source. 4. Expect environment credentials to stop working: under these keys `ANTHROPIC_API_KEY`, `ANTHROPIC_AUTH_TOKEN` and `apiKeyHelper` are blocked at startup, since organization membership can't be verified for them. Know the remaining gap: on the terminal's interactive login screen `forceLoginMethod` pre-selects a method without enforcing it, and Claude Code doesn't check which organization a Claude Console credential belongs to. ## Common misreadings - "ZDR means nothing about our usage is kept" — administrative data such as account emails, seat assignments and analytics metadata follows standard retention. - "ZDR makes our MCP integrations safe to send anything" — data a server receives is governed by that service, not by Anthropic's arrangement. - "Our new subsidiary organization inherits ZDR" — each organization is enabled separately. ## What ZDR does not change on the laptop ZDR is a server-side promise. Claude Code still writes plaintext session transcripts under `~/.claude/projects/` for `/resume`, kept for `cleanupPeriodDays` (30 days by default). Lowering that value, or setting `CLAUDE_CODE_SKIP_PROMPT_HISTORY`, is the local half of a retention policy. ## Summary ZDR is narrow by design — inference, one organization, Anthropic's own platform — and it only protects sessions that actually sign in to that organization. The managed login keys are what make that true for a fleet.

  • Why deliver the login keys through device management instead of only the claude.ai admin console?
    Server-managed settings are fetched only by sessions already authenticated into your organization, so they can't stop a developer's first login going to a personal account. Device-managed settings apply before any login. If both channels exist, set the keys in both, since managed sources don't merge and these keys aren't read across sources.
  • A developer on a ZDR organization routes Claude Code through Amazon Bedrock. Is that covered?
    Not by Anthropic's ZDR for Claude Code, which applies only to Anthropic's direct platform. Bedrock, Google Cloud's Agent Platform and Microsoft Foundry deployments follow those platforms' own retention policies, so the cloud provider's terms decide what is kept.

  • ZDR comes with every Enterprise plan and is switched on in admin settings.
  • ZDR also covers whatever an MCP server receives from Claude Code.
  • Enabling ZDR once covers every organization under the same account.
  • With ZDR on, nothing is ever retained, even after a policy violation.
  • ZDR stops Claude Code keeping transcripts on the developer's own machine.