Move project context from ChatGPT to Claude
Continue a ChatGPT project in Claude with a checked brief, source references, and a clear next task instead of a vague chat summary.
Move work between tools 3 min read
Claude memory import: what moves and what stays behind
Only what you paste moves. You run Anthropic’s export prompt in the old assistant, paste the reply into Claude, and Claude saves it as memory entries you can review. Your old chats, files, and anything the old tool learns later stay behind.
ON THIS PAGE
How the import works
Anthropic’s article gives you a prompt to run in the old assistant. It asks for every stored memory in one code block, with dates where they exist, then asks whether the list is complete. You can edit it to leave out anything sensitive.
Paste the reply into Claude’s import box and choose “Add to memory”, and Claude saves the key points as separate entries. Anthropic lists the import for Free, Pro, Max, and Team plans on the web and Claude Desktop. It calls the feature experimental, so some lines may not stick. Menu paths differ between Claude’s two memory layouts, so follow the article.
What stays behind
The import copies text once. Anything the old assistant leaves out cannot arrive, and no entry links back to the ChatGPT chat it came from.
- Chat history: ChatGPT also draws on past chats, and OpenAI says that context can change. The export is a list; the conversations stay in ChatGPT.
- Files and connected apps: files in ChatGPT’s Library and content from apps like Gmail do not travel with a pasted list.
- Non-work and sensitive details: Claude may not keep personal details unrelated to work. By default it also skips topics such as health, and it never saves items like government ID numbers.
- Later changes: a correction you make in ChatGPT next week stays there.
Check what Claude kept
After the import, Claude offers a new chat that asks what it learned. Compare its answer with your export, then open the memory list in Claude’s settings. In the current layout, memory sits under Topics, and you can edit or delete each one.
Keep one project’s detail out of general memory. Claude gives each project its own memory space, so a client’s pricing belongs in that project’s brief. For the exact wording of an old decision, go back to the original chat.
TAKE THIS WITH YOU
Mark up the export before you paste it
An illustrative template. Replace the details with material you are allowed to share.
Mark each line before you paste it: KEEP: true and useful PRUNE: a duplicate, or too personal STALE: fix it or drop it BRIEF: belongs to one project only KEEP [2026-03-02] - Prefers British spelling, short paragraphs KEEP [date unknown] - Wants the recommendation first PRUNE [2026-04-02] - Likes concise answers (same as line 1) STALE [2026-05-09] - Launch planned for June (moved to October) BRIEF [2026-06-11] - Client X pricing tiers and rollout order PRUNE [2026-07-01] - Asked about a sore knee once After the import, in Claude: - Open the memory list; fix wording, delete pruned lines. - Check each STALE line is corrected or gone. - Put BRIEF lines in that project’s brief.
USING THIS WITH DRAY
Keep the context. Check the source.
Dray does not sync with Claude’s own memory, but it can import the export files ChatGPT, Claude and Grok send you. Choose Import on your dashboard: your browser reads the file and sends Dray the text of your chats and the names of attached files. A Dray handoff opens Claude with a request to read a saved chat itself through Dray’s connector, so Claude works from the source instead of a list pasted once. It covers only chats Dray has saved.
Available sources and connections depend on your setup. Read the privacy policy before granting access.
Sources & next reads
Product details are grounded in the guides below. Example projects and prompts are illustrative, not customer results.