Import context
Bring your preferences and past project context over from another AI tool
If you have been using another AI assistant, you do not have to start from scratch. Cosmo can read an export from that tool and carry over two things: how you like to work, and what you were working on.
Where to find it
Click your name at the bottom-left of the sidebar and choose Import context. This works any time — whether you are brand new to Cosmo or have been using it for a while and are only now moving your history across.
What you can import
Upload a file, or paste the text straight in. Cosmo accepts:
- JSON exports, whatever their internal shape
- Markdown and plain text — including notes you wrote yourself
You do not need to tidy the file up first. Cosmo reads whatever prose it can find and ignores the machine fields around it.
Only files with a .json, .md, or .txt extension are accepted — anything else is rejected
before it is read. A file is also checked for whether it is genuinely text once picked; a file
renamed to look like one of these types but that is actually binary content is rejected too,
rather than being read as garbled text.
Files up to 15 MB can be uploaded. Of that, up to 2 MB of actual text is processed in one import — a very large archive is better imported one project or one date range at a time.
What Cosmo takes from it
Two kinds of thing, saved in two different places, because they are useful in different ways.
Preferences
Durable facts about how you work — that you like short answers, that you use ISO dates, that "TD" means Techdome, that you work on a travel booking platform.
These are kept short on purpose and are included in every conversation from then on. That is what makes Cosmo behave the way your previous tool did without you having to explain yourself again.
Projects
The substance of what you were working on — one document per project or topic.
These are saved as documents in Context, a space of their own in the Hub. They are not attached to every conversation, because that would be a lot of text to carry around. Instead Cosmo searches them when you refer to earlier work: ask "what did we decide about the payment provider?" and it will look, find the relevant document and read it.
Finding what you imported
Once you have imported at least one project, Context appears in the Hub sidebar. You can also jump straight there with View in Hub when an import finishes.
It is kept apart from your own workspaces on purpose — everything in it came from another tool and was extracted by Cosmo, rather than written by you — and it shows what an ordinary workspace can't:
- Where each document came from — "From ChatGPT", "From Claude" — and when it was imported.
- A one-line summary of each project, so you can tell them apart without opening them.
- Updated by re-import on a document a later import refreshed. Importing the same project again updates its document in place rather than adding a copy, and this is how you can see that it happened.
- At a glance: how many documents you have imported, and when you last imported anything.
You can search by title, summary or source, and open, read or download any document. To remove one you no longer want Cosmo to draw on, open it and delete it. To import more, use Import context from your profile menu.
While you are looking at Context, the chat is focused on it, so a question like "what did I decide about payments?" is answered from your imported documents. Anything Cosmo notes down while you work there is kept under Memory at the bottom of the page, where you can read, edit or delete it — the same as on any workspace.
Cosmo finds these documents both by title and by what's actually written inside them, so a question does not have to echo a title to find its answer. A clear title still helps — it is what Cosmo shows you when it names where an answer came from, and it is what keeps a document visible in your always-on memory as you import more over time (see below).
You review everything first
Nothing is saved until you say so. After Cosmo reads your file you get a list of everything it found, and you can:
- uncheck anything you would rather it did not keep
- edit the wording of any preference
- rename any project document
- choose whether preferences apply to just you, to a team you belong to, or to everyone in your organization
Preferences default to just you. An import is your own accumulated context, so it is not applied to your colleagues unless you choose that deliberately.
If part of your file could not be read — even after Cosmo tried it a second time — the review says so above the list. Everything that is listed is complete and can be saved as usual; importing the same file again may pick up what was missed.
The team option appears only if you belong to a shared team — your personal team is not listed, since saving there would be the same as saving to yourself.
This review step is not just a convenience. Anything in an imported file is treated as information about you, never as instructions to follow — and reading the list before it is saved is what makes sure of it. If your export happens to contain text that looks like an instruction, you will see it in the list and can remove it.
If a file appears to contain content specifically designed to interfere with how Cosmo reads it, the import is refused outright with an explanation, before anything is processed. This is a narrow, high-confidence check rather than a general filter on suspicious-sounding text — an ordinary sentence is never rejected because of this.
Importing again later
Running an import a second time replaces what the previous one from the same tool put there, rather than adding a second copy. The same is true of the project documents: a project you import twice is updated, not duplicated.
Anything you have written in your own memory files is left alone.
As you import more over time
Cosmo keeps a short, always-on list of your most recently imported or updated project documents, so it can recognize when a question touches one of them without you having to name it. That list only has room for so many entries — once you have imported enough projects, the oldest ones drop off it.
Dropping off the list does not mean a document is lost or stops being found. Every imported document stays in Context and stays fully searchable — by title and by what's written inside it — whenever Cosmo looks there. The list is just a shortcut so recent work surfaces without a search; older work is still one question away.
What it does not do
Cosmo does not connect to your other AI tool directly — there is no login, and no data is fetched on your behalf. You export from that tool and hand Cosmo the file, which means you can see exactly what you are sharing before you share it.
Retrieval is also not quite the same as having been in the original conversation. Cosmo has to search for the right document rather than already knowing it, so results can occasionally miss something a human would have caught instantly. Giving your project documents clear names during review still helps — a good title is the fastest way for both you and Cosmo to recognize the right document.