Information about Anthropic's standard retention policies is set out in Anthropic's commercial data retention policy and consumer data retention policy.
When users use API endpoints with zero data retention (ZDR), customer data submitted through those endpoints is not stored at rest after the API response is returned except where needed to comply with law or combat misuse. Subject to these exceptions, when using ZDR-enabled endpoints, customer data is processed in real time and promptly discarded, with no logging or non-ephemeral storage of prompts or outputs.
Different APIs and features have different storage and retention needs. Where an API or feature doesn't require storage of customer prompts or responses, it may be eligible for ZDR. Where an API or feature necessarily requires storage of customer prompts or responses, Anthropic designs for the smallest possible retention footprint. For these features:
In the ZDR eligibility table, some features are marked "Yes (qualified)" in the ZDR eligible column. If your organization has a ZDR arrangement, you can use these features with confidence that what Anthropic retains is narrow and is required for optimal performance.
What ZDR covers
What ZDR does NOT cover
For the most up-to-date information on what products and features are ZDR-eligible, refer to your contract terms or contact your Anthropic account representative.
| Feature | Endpoint | ZDR eligible | Data retention details |
|---|---|---|---|
| Messages API | /v1/messages | Yes | Standard API calls for generating Claude responses. |
| Token counting | /v1/messages/count_tokens | Yes | Count tokens before sending requests. |
| Web search | /v1/messages (with web_search tool) | Yes1 | Real-time web search results returned in the API response. |
| Web fetch | /v1/messages (with web_fetch tool) | Yes1 2 | Fetched web content returned in the API response. |
| Memory tool | /v1/messages (with memory tool) | Yes | Client-side memory storage where you control data retention. |
| Context management (compaction) | /v1/messages (with context_management) | Yes | Server-side compaction results are returned/round-tripped statelessly through the API response. |
| Context editing | /v1/messages (with context_management) | Yes | Context edits (tool use clearing + thinking clearing) are applied in real time. |
| Fast mode | /v1/messages (with speed: "fast") | Yes | Same Messages API endpoint with faster inference. ZDR applies regardless of speed setting. |
| 1M token context window | /v1/messages | Yes | Extended context processing uses the standard Messages API. |
| Adaptive thinking | /v1/messages | Yes | Dynamic thinking depth uses the standard Messages API. |
| Citations | /v1/messages | Yes | Source attribution uses the standard Messages API. |
| Data residency | /v1/messages (with inference_geo) | Yes | Geographic routing uses the standard Messages API. |
| Effort | /v1/messages (with effort) | Yes | Token efficiency control uses the standard Messages API. |
| Extended thinking | /v1/messages (with thinking) | Yes | Step-by-step reasoning uses the standard Messages API. |
| PDF support | /v1/messages | Yes | PDF document processing uses the standard Messages API. |
| Search results | /v1/messages (with search_results source) | Yes | RAG citation support uses the standard Messages API. |
| Bash tool | /v1/messages (with bash tool) | Yes | Client-side tool executed in your environment. |
| Text editor tool | /v1/messages (with text_editor tool) | Yes | Client-side tool executed in your environment. |
| Computer use | /v1/messages (with computer tool) | Yes | Client-side tool where screenshots and files are captured and stored in your environment, not by Anthropic. See Computer use. |
| Fine-grained tool streaming | /v1/messages | Yes | Streaming tool parameters uses the standard Messages API. |
| Prompt caching | /v1/messages | Yes | Your prompts and Claude's outputs are not stored. KV cache representations and cryptographic hashes are held in memory for the cache TTL and promptly deleted after expiry. See Prompt caching. |
| Structured outputs | /v1/messages | Yes (qualified) | Your prompts and Claude's outputs are not stored. Only the JSON schema is cached, for up to 24 hours since last use. See Structured outputs. |
| Tool search | /v1/messages (with tool_search tool) | Yes (qualified) | Only tool catalog data (names, descriptions, argument metadata) is stored server-side. See Tool search. |
| Batch processing | /v1/messages/batches | No | 29-day retention; async storage required. See Batch processing. |
| Code execution | /v1/messages (with code_execution tool) | No | Container data retained up to 30 days. See Code execution. |
| Programmatic tool calling | /v1/messages (with code_execution tool) | No | Built on code execution containers; data retained up to 30 days. See Programmatic tool calling. |
| Files API | /v1/files | No | Files retained until explicitly deleted. See Files API. |
| Agent skills | /v1/messages (with skills) / /v1/skills | No | Skill data retained per standard policy. See Agent skills. |
| MCP connector | /v1/messages (with mcp_servers) | No | Data retained per standard policy. See MCP connector. |
1 Dynamic filtering is not ZDR-eligible.
2 While web fetch is ZDR-eligible, website publishers may retain request data (such as fetched URLs and request metadata) according to their own policies.
Cross-Origin Resource Sharing (CORS) is not supported for organizations with ZDR arrangements. If you need to make API calls from browser-based applications, you must:
Even with ZDR arrangements in place, Anthropic may retain data where required by law or to combat Usage Policy violations and malicious uses of Anthropic's platform. As a result, if a chat or session is flagged for such a violation, Anthropic may retain inputs and outputs for up to 2 years.
Was this page helpful?