September 17, 2026 · 6 min read

Exporting Your Claude Preferences to ChatGPT and Gemini

Q: What files does exporting to Gemini produce?

One file, gemini-instructions.txt, covering your background, current focus, and response style as plain text. Paste its contents into Gemini's Settings → Saved Info.

open-context's Export tab can take the same preferences profile you built for Claude and turn it into ChatGPT's two custom-instruction fields or a single Gemini Saved Info document — in seconds, entirely in your browser. You fill in one form once; the open-context exporter does the vendor-specific formatting for you.

Key takeaways

open-context's VendorExporter pattern translates one UserPreferences profile into Claude, ChatGPT, or Gemini format — no separate profile per vendor.
Exporting to ChatGPT produces two files, chatgpt-what-to-know.txt and chatgpt-how-to-respond.txt, matching OpenAI's actual two custom-instruction fields.
Exporting to Gemini produces one file, gemini-instructions.txt, formatted for Gemini's Settings → Saved Info.
None of the three exporters move full conversation transcripts — only preference text, plus a short line of recent conversation titles if you've imported any.
The whole export step runs client-side with Copy and Download buttons per file — nothing is uploaded anywhere.

The direction most tools don't build for

Most AI migration tools move in one direction: into the vendor everyone happens to be switching to. open-context's own ChatGPT-to-Claude guide is that kind of guide. But a preferences profile — your role, your tone preference, your tech stack, what you're currently working on — isn't Claude-specific. It's just a description of you. There's no reason it should only ever flow toward one vendor.

The Export tab in open-context's web UI treats it that way. Once you've filled in the Preferences form, you can generate output for Claude, ChatGPT, or Google Gemini from the same data, and switch between them without re-entering anything.

One profile, three export targets

Every vendor in open-context implements the same small interface: a preferences formatter and a conversations formatter, each returning a list of named text files. The UI just asks the exporter for the currently selected vendor to format the current profile:

interface VendorExporter {
  info: VendorInfo
  exportPreferences(preferences: UserPreferences): ExportResult
  exportConversations(conversations, preferences): ExportResult
}

The UserPreferences object underneath is the same one the Preferences editor writes to — communication style (tone, detail level, whether you want code examples or step-by-step answers), a technical profile (languages, frameworks, tools, experience level), work and personal context, current focus, and any free-form custom instructions you've added. Each exporter reads those same fields and writes them out in the shape its target vendor expects.

What gets generated for ChatGPT

ChatGPT's own personalization settings ask two separate questions — "What would you like ChatGPT to know about you?" and "How would you like ChatGPT to respond?" — under Settings → Personalization → Custom Instructions (OpenAI Help Center). open-context's ChatGPT exporter mirrors that split exactly, generating:

File	Maps to
`chatgpt-what-to-know.txt`	"What would you like ChatGPT to know about you?"
`chatgpt-how-to-respond.txt`	"How would you like ChatGPT to respond?"

The first file pulls your role, industry, tech stack, background, current projects and goals, and — if you've imported conversations — the titles of your three most recent ones. The second file translates your tone, detail level, formatting, and behavior preferences (follow-up questions, risk warnings, proactiveness) into plain instruction sentences, then appends any custom instructions you wrote verbatim. Per OpenAI's help center, free and Go accounts cap custom instructions at 1,500 characters per field, while Plus, Pro, Business, Enterprise, and Education accounts get 5,000 — worth checking the generated file length against your plan before pasting.

What gets generated for Gemini

Gemini's equivalent is its Saved Info settings, where Gemini stores standing details and instructions it references across conversations (Gemini Apps Help). Unlike ChatGPT's two-field split, open-context's Gemini exporter writes a single gemini-instructions.txt with four labeled sections — About me, Current focus, Response style, and Additional instructions — built from the same preferences data. It's meant to be pasted in as one block rather than split across fields.

Step-by-step: exporting from open-context

Fill in Preferences. Open the Preferences tab and complete the sections that apply to you — this is the profile every export pulls from.
Optionally import conversations. If you've imported a ChatGPT or Claude export, select the conversations you want reflected in the "recent topics" line.
Open Export and pick a vendor. Choose Claude, ChatGPT, or Gemini from the vendor grid and click Generate.
Copy or download each file. Every generated file has its own Preview, Copy, and Download controls, plus a Download All button for the full set.
Paste into the vendor's settings. ChatGPT: Settings → Personalization → Custom Instructions. Gemini: Settings → Saved Info.

What this doesn't do

It's worth being precise about the limits. open-context can't read memory out of your live Claude, ChatGPT, or Gemini account — there's no API for that, and the tool never connects to any vendor account at all. What it exports is the profile you built inside open-context itself, which happens to be the same one you'd hand to Claude. And none of the three exporters — Claude's included — support full conversation import into the target vendor; every VendorInfo in the exporter registry sets supportsConversationImport: false. Recent conversation titles get folded into the text as one line of context, not full transcripts. For moving actual conversation content, see Claude Preferences vs Memory for how the markdown conversation files fit into that separate picture.

Build your preferences profile once, export it anywhere.

Try open-context's Export tab →

FAQ

Does open-context read my saved memory out of my Claude account?

No. open-context never connects to any vendor's account. It exports the same preferences profile you filled in inside open-context — the one you'd use for Claude — into ChatGPT and Gemini's own instruction formats. There is no API for reading Claude's account-side memory, so this works off your local profile, not your live Claude chat history.

Do I need to import ChatGPT or Gemini history first?

No. The ChatGPT and Gemini exports only need the Preferences form filled in. Importing conversations is optional — if you have some imported, their titles get added as a short "recent topics" line, but no conversation content is included.

Does this export my full conversation history into ChatGPT or Gemini?

No. All three built-in exporters — Claude, ChatGPT, and Gemini — set supportsConversationImport to false. They only produce preference and memory text, never full conversation transcripts.

What files does exporting to Gemini produce?

One file, gemini-instructions.txt, covering your background, current focus, and response style as plain text. Paste its contents into Gemini's Settings → Saved Info.

Is my data sent anywhere during the export?

No. The export runs in the browser tab you're already using — it formats your local preferences state into text and offers Copy and Download buttons. Nothing is uploaded to open-context or any third party.