Memory Portability And Attempted Full-Memory Export

Issue 60 Edition 2026-03-01 5 min read

Not accepted General

Sources: 1 • Confidence: Medium • Updated: 2026-04-13 03:48

Key takeaways

The prompt instructs the model to output every stored memory and learned context in a single code block formatted as "[date saved, if available] - memory content" without summarizing, grouping, or omitting entries.
Anthropic's "import your memories to Claude" feature at claude.com/import-memory is implemented as a prompt.
The prompt defines the requested coverage of memories/context to include response-style instructions, personal details, projects/goals, tools/frameworks, preferences/corrections, and any other stored context, with verbatim preservation where possible.
The text requests a complete export of stored memories because the user is moving to another service.

The prompt instructs the model to output every stored memory and learned context in a single code block formatted as "[date saved, if available] - memory content" without summarizing, grouping, or omitting entries.
The prompt defines the requested coverage of memories/context to include response-style instructions, personal details, projects/goals, tools/frameworks, preferences/corrections, and any other stored context, with verbatim preservation where possible.
The text requests a complete export of stored memories because the user is moving to another service.

Anthropic's "import your memories to Claude" feature at claude.com/import-memory is implemented as a prompt.

Does the target system return stored-memory records verbatim when instructed, or does it refuse/limit outputs to prevent full-memory dumps?
What is the boundary between "stored memories" and "learned context" in the system, and which of the requested categories are actually persisted versus inferred at response time?
What access controls, consent flows, auditing, or policy checks (if any) gate memory import/export behavior for this feature?
Is there a supported, system-level export capability (as opposed to a user-authored prompt) that reliably enables portability across AI services?
Is there any direct operator, product, or investor decision readthrough evidenced in this corpus beyond the existence/description of the prompt and user intent?

Prompt-based memory import and export could indicate a lighter-weight portability feature, potentially reducing engineering effort but increasing reliance on instruction-following. This may create product and trust implications if users can coerce broader disclosure than intended.
User demand for memory portability suggests emerging expectations for cross-service portability. If broadly adopted, it could shift competitive dynamics toward standardized export formats and consent flows, benefiting platforms that offer compliant, auditable portability.

Product documentation or demos showing memory import or export is implemented via prompts rather than a hardened system workflow, including the exact user flow and guardrails.
Evidence of access controls such as explicit consent, scoped export, audit logs, and limits that prevent full memory dumps, alongside clarified definitions of stored memories versus learned context.
Adoption signals such as public usage metrics, enterprise feature announcements, or policy updates that position memory portability as a differentiated capability.

Testing or disclosures showing the system refuses or heavily limits any attempt to output stored memories verbatim, indicating the prompt described does not work in practice.
Clarification that the import feature is not prompt-mediated but uses a controlled backend pipeline, reducing the implication that a user can extract memories through instruction alone.
Policy or product changes that disable portability workflows or restrict them to narrow categories, indicating low strategic importance or high risk.