Zero Data Retention Planning for Claude Code Enterprise

TL;DR

Zero data retention planning maps contractual promises to real Claude Code data paths: model requests, logging, analytics, MCP tools, and integrations. Inventory flows, verify enterprise settings, block non-compliant MCP servers, and document verification steps before declaring rollout complete.

Prerequisites & Requirements

• {"task": "ZDR docs reviewed", "description": "Official zero-data-retention documentation is read and summarized"}
• {"task": "Data path inventory", "description": "Prompts, tools, logs, analytics, Slack, and MCP flows are listed"}
• {"task": "Stakeholders identified", "description": "Legal, security, and platform engineering sign-off owners are named"}
• {"task": "Pilot tenant ready", "description": "Verification runs on non-production users first"}
• {"task": "Evidence folder created", "description": "Config exports and test logs will be stored for audits"}

Core Concepts Explained

ZDR is a program, not a checkbox

Contracts define retention exclusions; engineering must map those promises to each integration that touches prompts or tool output.

Local and remote retention differ

Claude Code ZDR addresses provider-side retention; MCP vendors, Slack, and internal log aggregators need separate review.

MCP expands the boundary

Every approved MCP server is a potential retention point outside core ZDR docs.

Verification needs evidence

Keep screenshots, config exports, and test results showing ZDR settings active for the deployment profile.

Step-by-Step Implementation Guide

1. Read official ZDR documentation. Capture supported configurations and explicit exclusions.

2. Inventory data paths. List prompts, tool I/O, session exports, analytics, Slack, and MCP flows.

3. Classify components. Mark in-scope ZDR, out-of-scope third party, or blocked pending review.

4. Configure enterprise settings. Apply managed settings aligned with ZDR requirements.

5. Block risky MCP defaults. Publish allowlists for MCP servers compatible with retention policy.

6. Run pilot verification. Execute scripted tasks and confirm no prohibited retention surfaces appear in logs or vendor dashboards.

7. Train champions. Teach intake redaction and MCP request procedures under ZDR.

8. Schedule re-verification. Re-run inventory when enabling Slack, analytics, or new MCP integrations.

ZDR Verification Checklist

• {"task": "Core ZDR settings enabled", "description": "Enterprise deployment profile matches documentation"}
• {"task": "MCP allowlist published", "description": "Only reviewed servers are approved for production"}
• {"task": "Analytics scope verified", "description": "Dashboards do not retain prohibited prompt content"}
• {"task": "Audit evidence stored", "description": "Dated exports and pilot logs are archived"}

Troubleshooting

Security questionnaire asks about MCP retention

Provide separate MCP vendor assessments; core ZDR docs may not cover them.

Analytics appears to store prompts

Verify whether analytics is disabled or anonymized per enterprise policy.

Developers use personal MCP servers

Enforce managed MCP policy and block unapproved servers in enterprise builds.

Audit requests evidence

Maintain dated config exports and pilot test logs prepared during initial rollout.

Duplicate Check

This guide is distinct from healthcare-hipaa-guide.mdx and financial-services-guide.mdx, which cover regulated industry workflows. This entry focuses on zero data retention planning mechanics for enterprise Claude Code.

References

• Claude Code zero data retention - https://code.claude.com/docs/en/zero-data-retention
• Enterprise network config - https://code.claude.com/docs/en/network-config
• Usage analytics - https://code.claude.com/docs/en/analytics