Migrating From Google Gemini to Claude: What Transfers

open-context can't import a Google Gemini chat export yet — that's an open roadmap item, not a supported path today. What it can do right now is the reverse: convert your Claude preferences and memory into a Gemini-formatted custom-instructions file, so the two assistants stay in sync while you decide where to land. This guide covers exactly what works today, what a Google Takeout export of your Gemini activity actually contains, and a model-agnostic way to carry context between assistants without waiting on either importer.

What actually transfers from Gemini to Claude today

If you're looking for a one-command way to pull your Gemini chat history into Claude the way open-context migrates ChatGPT exports, it doesn't exist yet. Google Gemini support is listed as a planned, unchecked item on the open-context roadmap — the parser that would read a Gemini export and normalize it into Claude-ready markdown hasn't been written.

What does transfer today runs in the other direction. open-context's export pipeline already supports writing your Claude preferences and memory out to a Gemini-compatible format, alongside Claude and ChatGPT. If you've built up preferences in open-context — communication style, technical profile, current projects — you can export a gemini-instructions.txt file today and paste it into Gemini's saved info or a custom Gem, even though the reverse (Gemini → open-context) isn't built yet.

Why there's no Gemini import parser yet

open-context's conversion pipeline is deliberately split into three stages — parse, normalize, generate — with one parser module per source provider. ChatGPT's parser reads conversations.json and walks its message tree into a common schema; a Gemini parser would do the same against Google Takeout's Gemini Apps activity format. The architecture already supports adding it as a self-contained module without touching the rest of the pipeline — it just hasn't been built yet, in part because Gemini's export isn't structured as a per-conversation file the way ChatGPT's is (more on that below).

Exporting Claude context into Gemini format today

The Gemini exporter that does exist takes your preferences.md-equivalent data — role, industry, technical profile, current projects, response-style settings, and any custom instructions you've written — and formats it as plain text sections Gemini understands: an "About me" block, a "Current focus" block, and a "Response style" block.

In practice, this makes open-context useful for keeping Gemini's custom instructions roughly aligned with whatever you've already set up in Claude, even while full Gemini chat import is still pending.

What's actually inside a Google Takeout Gemini export

If you want to look at your own Gemini history in the meantime, Google Takeout can produce it — you just have to know where to look. Select My Activity, not the top-level "Gemini" category (that one only covers saved Gems), then narrow it to Gemini Apps and set the output format to JSON instead of the HTML default. The result is an activity log — typically something like Takeout/My Activity/Gemini Apps/MyActivity.json — containing your prompts, Gemini's responses, and timestamps for each interaction.

That's meaningfully different from ChatGPT's export. ChatGPT's conversations.json is a purpose-built conversation tree with stable per-conversation IDs, which is exactly what open-context's parser walks today. A Takeout activity log is a flatter, more general-purpose record of "things you did in Google products" — usable, but it needs a dedicated parser to turn into the same normalized schema, which is the work the roadmap item covers.

A model-agnostic alternative: MCP persistent memory

If the goal isn't a one-time import but rather "I don't want to keep re-explaining myself to whichever assistant I'm using," open-context's MCP server sidesteps the whole import question. Once connected to Claude Code or Claude Desktop, its save_context and recall_context tools let Claude write and retrieve notes from a local store — ~/.opencontext/contexts.json by default — that you build up going forward, independent of any provider's export format. It doesn't recover your Gemini history retroactively, but it means you stop losing new context the moment you set it up.

Contributing a Gemini import parser

open-context is open source, and the pattern for adding a source is documented: study the provider's export format, write a parser under src/parsers/ that maps its roles and content into the normalized schema, add anonymized fixture data, and register it in the CLI. A Gemini parser against the Takeout JSON format described above is exactly the kind of contribution the project is structured to accept.

FAQ

Can open-context import my Google Gemini chat history today?

Not yet. open-context currently parses full conversation history only from ChatGPT exports. Gemini import is a tracked roadmap item, not a shipped feature.

Can I export my Claude preferences to use with Gemini?

Yes. open-context's export pipeline already includes a Gemini exporter that converts your preferences and memory into a Gemini-formatted custom-instructions text file, ready to paste into Gemini's saved-info settings or a custom Gem.

What does a Google Takeout export of Gemini activity contain?

If you select My Activity, then Gemini Apps, and set the format to JSON, Google Takeout gives you a JSON activity log with your prompts, Gemini's responses, and timestamps. It's a raw activity export, not a structured conversation file the way ChatGPT's conversations.json is.

Is Gemini import going to be added to open-context?

It's on the public roadmap. open-context's parser layer is built so each provider is a self-contained module, so adding Gemini support means writing one new parser against the Takeout JSON schema rather than reworking the pipeline.

Does any of this send my Gemini or Claude data to an external server?

No. Every part of open-context — the CLI, the web UI, the exporters, and the MCP server — runs on your own machine. Nothing described here uploads your conversations anywhere.