Claude Import Memory
One-copy migration that ports ChatGPT/Gemini context into Claude’s Memory
Claude Import Memory is the fastest choice for power users and teams who need to port working context from ChatGPT/Gemini into Claude Memory. In LinkStart Lab, the best results came from importing a tightly-scoped “work profile” (tools, tone, formats) rather than raw chat dumps. It shines for onboarding and continuity, but you still need human review to prevent noisy or outdated preferences from being saved.
Why we love it
- Turns provider lock-in into a one-copy workflow: prompt → paste → editable memory edits
- Practical quality control: review entries via Manage edits before they shape future outputs
- Great for repeatable systems (sales outreach, code review style, meeting-note templates) where consistency matters
Things to know
- Requires a paid Claude plan; Free users can’t rely on Memory imports as a migration baseline
- Imports are experimental, and full propagation can take up to 24 hours in some cases
- Not a true API-based sync; it’s a copy/paste pipeline, so secret hygiene is on you
About
Claude Import Memory is Anthropic’s built-in migration flow that extracts your working preferences, projects, and recurring context from other AI assistants and turns them into editable Memory entries inside Claude. It is designed for fast onboarding: you paste a purpose-built prompt into ChatGPT, Google Gemini, or another provider, then paste the results into Claude’s Settings → Capabilities → Memory to “Add to memory.” Claude will extract key items into individual memory edits you can review and adjust, and imported changes can take up to 24 hours to fully appear. Claude offers a free plan, with paid tiers starting at $20/month. Import Memory requires a paid plan. It is less expensive than most enterprise-grade migration or RPA workarounds for “context portability.”
Key Features
- ✓Extract context with a dedicated prompt to reduce manual re-onboarding
- ✓Convert pasted exports into editable, per-item Memory entries
- ✓Review and adjust imported memory edits for accuracy and safety
- ✓Bootstrap consistent outputs across sessions for faster workflows
Frequently Asked Questions
No—Import Memory requires a paid Claude plan. Pricing starts at Claude Pro ($20/month), with higher-usage options like Max ($100–$200/month) and team plans (Team Standard from $25/seat/month). The key point: you can have a free Claude account, but the Memory feature that Import uses is gated behind paid tiers.
The core difference is portability vs native accumulation. While ChatGPT Memory is optimized for staying inside the OpenAI ecosystem, Claude Import Memory has an absolute advantage for migration speed because it’s explicitly built as a prompt → paste pipeline that converts exports into editable “memory edits” inside Claude. If your KPI is “time to first productive Claude session,” Import Memory usually wins because it reduces re-onboarding to a single copy/paste plus a quick review pass.
Use a narrow, work-only “profile” export, not a full chat-history dump. While huge exports maximize recall, Claude Import Memory performs better when entries are explicit, atomic, and scoped (e.g., “Preferred output: Markdown tables,” “Tech stack: Python + PostgreSQL,” “Do: cite assumptions,” “Don’t: use emojis”), because Claude stores them as per-item memory edits you can delete or edit later. The practical workaround for “memory weirdness” is to import 30–80 high-signal lines first, validate output behavior, then expand.
You start it in Claude under Settings → Capabilities → Memory by choosing “Start import,” then pasting your exported text and clicking “Add to memory.” Claude turns the paste into multiple memory edits and lets you review them via “Manage edits.” Expect full visibility within 24 hours in some cases, so don’t judge success only by the very next message.
It can be secure if you treat the import as sensitive data handling. The import is literally a copy/paste of content into Claude Memory, and you can (and should) customize the export prompt to exclude secrets before importing; avoid API keys, passwords, private customer data, and proprietary source code. The safest workflow is to import only work preferences and public facts, then add any sensitive details later as controlled, minimal edits—or keep them out of Memory entirely.
Yes—this is one of the best uses of Import Memory. Create a compact ruleset in your source assistant (e.g., “Use black + ruff,” “Prefer SQL migrations,” “PostgreSQL naming: snake_case,” “Use pgvector for embeddings,” “Return diffs in unified format”), export it as a single block, and import it so Claude stores each rule as an editable memory edit. While a wiki can document standards, Import Memory gives Claude the defaults so your code reviews, SQL snippets, and refactors start aligned without re-prompting every time.