Hermit vs Claude's Memory Import
Two ways to bring your ChatGPT context into Claude. One reads ~40 stored memories. The other reads your entire conversation history.
| Feature | Claude Memory Import | Hermit |
|---|---|---|
| What it reads | ChatGPT's stored memories (~40-50 items) | Your full conversations.json export (all conversations) |
| Output size | ~7,500 characters | ~110,000 characters — 14x more context* |
| Output type | Flat list of facts | Structured profiles with behavioral instructions |
| Temporal awareness | No — stale facts stay as current | Yes — ACTIVE/PAST labels on all items |
| Works with | Claude only | Claude, Gemini, any LLM |
| Per-topic depth | No — one flat list | Yes — one profile per project/theme |
| Setup time | ~2 minutes | ~5 minutes (upload + processing) |
| Price | Free | Free stats. Profiles from €19.90 |
What it reads
Claude Memory Import
ChatGPT's stored memories (~40-50 items)
Hermit
Your full conversations.json export (all conversations)
Output size
Claude Memory Import
~7,500 characters
Hermit
~110,000 characters — 14x more context*
Output type
Claude Memory Import
Flat list of facts
Hermit
Structured profiles with behavioral instructions
Temporal awareness
Claude Memory Import
No — stale facts stay as current
Hermit
Yes — ACTIVE/PAST labels on all items
Works with
Claude Memory Import
Claude only
Hermit
Claude, Gemini, any LLM
Per-topic depth
Claude Memory Import
No — one flat list
Hermit
Yes — one profile per project/theme
Setup time
Claude Memory Import
~2 minutes
Hermit
~5 minutes (upload + processing)
Price
Claude Memory Import
Free
Hermit
Free stats. Profiles from €19.90
*Based on a 1,258-conversation export over 3 years. Results scale with your usage — a 100-conversation export typically produces 3-5x more context than Claude's Memory Import.
How Claude's Memory Import works
Anthropic built an official way to import your ChatGPT context into Claude. You go to claude.ai/import-memory, copy a prompt, paste it into ChatGPT, and ChatGPT dumps its stored memories about you. You then paste that output back into Claude's Memory settings.
It works well for a quick transfer. The limitation is that ChatGPT's “memories” are a curated subset — typically 40-50 facts that ChatGPT chose to save over time. They have no structure, no temporal markers (so a job you left in 2024 might still say “works at”), and no behavioral instructions for Claude to follow.
For users with fewer than 100 conversations or who mainly need basic preferences transferred, this is a solid option.
How Hermit processes your ChatGPT export
Hermit takes a different approach. Instead of reading ChatGPT's stored memories (which are a fraction of what ChatGPT actually knows), Hermit reads your entire conversation export — every conversation you've ever had.
- You export your data from ChatGPT (Settings → Data Controls → Export)
- Upload your conversations.json to Hermit
- A 6-step AI pipeline processes your history:
- Each conversation is scored for relevance (most one-shot queries are filtered out)
- Conversations are clustered by project and theme
- Each cluster gets a structured profile with behavioral instructions
- A Global Profile synthesizes your patterns, philosophy, and working style
- A current snapshot captures your last 6 months
- Memory bullets are generated for Claude's memory system
The output is a ZIP file with everything structured and ready to import — into Claude Projects, Claude Memory, Gemini Gems, or any LLM that accepts system prompts.
When to use which — ChatGPT to Claude migration guide
Use Claude's Memory Import if:
- •You have fewer than ~100 ChatGPT conversations
- •You mainly need basic preferences transferred (tone, formatting, tools you use)
- •You want the fastest possible setup (2 minutes, free)
- •You only use Claude (don't need portability)
Use Hermit if:
- •You have months or years of ChatGPT history
- •You want Claude to understand how you think, not just what you've told it
- •You need temporal awareness (what's current vs. what's past)
- •You use multiple LLMs (Claude + Gemini + others)
- •You want per-project/per-topic context (not one flat list)
- •You've been a ChatGPT power user and want to preserve that depth
What the output looks like
Claude Memory Import output
[2024-03] — The user is a product manager [2024-06] — The user is training for a half-marathon [2024-09] — The user had a teeth whitening appointment [2024-11] — The user uses Supabase and React Native [unknown] — The user prefers concise answers [unknown] — The user has a cat named Pixel [unknown] — The user drinks coffee every morning [unknown] — The user prefers dark mode in all apps ... (~35 more facts)
Hermit output (Global Profile excerpt)
You are a product designer who thinks in systems. Your career trajectory — from agency UX to in-house PM to solo founder — reflects a pattern of seeking increasing autonomy over the problems you solve. You process decisions by writing them out, often in conversation with AI, and you trust structured frameworks over intuition alone. When you ask for recommendations, lead with a single clear pick — you dislike balanced option menus. When discussing your startup, anchor to the real estate market; you have deep domain knowledge of French notary processes. ACTIVE: Building MVP (React Native + Supabase) ACTIVE: Training for Paris Marathon — April 2026 PAST: PM job search at healthtech startups (concluded Dec 2024)
The examples above are simplified excerpts. Here's a complete profile with all three outputs:
What Hermit builds from your conversations
Here's a real profile — generated from someone's ChatGPT export, shared with their permission.
Claude Memory Import
Recommended first stepPaste once into Claude's Memory Import. Claude knows you everywhere, across every conversation.
How to work with this person
Lead with visual references when explaining abstract concepts — she thinks in images and spatial relationships before she thinks in words.
and 5 other behavioral instructions
Who she is
Maria is a 31-year-old French product designer working in the Paris startup ecosystem — someone who came up through visual design but now operates at the intersection of UX strategy, brand systems, and front-end implementation. She is a senior designer at a Series B fintech, where she owns the design system and leads a two-person design team, but spends her evenings and weekends building toward independence: freelance clients, a side project (a collaborative moodboarding tool), and a growing LinkedIn presence where she writes about design decisions with unusual honesty.
Your thinking is shaped by a fine arts education (École Estienne) that gave you typographic and editorial instincts, followed by a pivot into product design that forced you to reconcile aesthetics with conversion metrics. You don't see these as opposed — you see most designers as having given up on the synthesis too early.
You are not a casual user of AI. You use ChatGPT as a design thinking partner — not to generate mockups, but to stress-test product reasoning, explore naming systems, structure client proposals, and think through career decisions out loud. You treat AI the way some designers treat sketchbooks: a space for thinking that is private, iterative, and accumulative.
Frequently asked questions
Ready to see what ChatGPT actually knows about you?
Upload your export and get free stats in minutes. Profiles from €19.90.
Get started for free