If you work across languages, you already know the small frictions that add up. The client meeting was in Spanish, but your project notes are in English because that's what your team reads. The supplier sent a quote in French, and you're not sure whether to translate it now or later. You captured a thought in your first language while walking, and now you have to decide whether to re-write it in English before the rest of your team can act on it.
Most note apps quietly assume a single working language. They sort, search, and surface notes as if everyone you collaborate with reads the same thing. For people who actually live in two or three languages a day, that assumption breaks the system in subtle ways — important information gets lost not because it wasn't captured, but because it was captured in the wrong language for the reader who needed it.
The fix is a vault that's genuinely language-agnostic about capture and retrieval, and an agent that can shift register and language when you ask. (For the broader case for capturing thoughts in whatever shape they arrive, see The Capture Habit: Remembering the Things That Actually Matter.)
Capture in whichever language is closest to the moment
The most important thing about a note is that it gets written down at all. The second most important thing is that it gets written down accurately. Forcing yourself to translate while you capture costs both speed and accuracy.
In Docapybara, you can write a page in any language. Notes are stored as plain markdown text, so there's no schema, no language tag, no required structure. A meeting note in Portuguese sits next to one in English next to one in Mandarin, and the vault doesn't care.
This matters more than it sounds. The friction of "should I capture in English or my first language?" disappears, which means more of the actual thought makes it to the page. The note from the Spanish-language client meeting can be in Spanish, with whatever code-switching is natural — and you can deal with the English version later, when there's a specific reader who needs one.
Search across the whole vault, not just one language
The problem with multilingual notes is that traditional search is keyword-based and language-specific. Search for "reunión" and you get the Spanish notes. Search for "meeting" and you get the English ones. The note from October that you can't quite remember the language of? You search both terms and hope.
The agent in Docapybara reads across the whole vault when you ask a question. If you ask "What did we agree about pricing with the supplier in October?" — in English — the agent finds the relevant note even if it was written in French, German, or Japanese. It pulls the relevant section and tells you which page it came from.
This is the move that makes a multilingual vault practical. You no longer have to remember which language a piece of information was captured in. You ask a question in whichever language is convenient, and the agent searches across everything.
Audio and meetings — speaker labels survive language switches
A lot of cross-language work happens in meetings where the conversation switches mid-sentence. A call between a French team and an Italian client might run mostly in English with French side-conversations and Italian asides for clarification. Trying to take notes in real time on that kind of call is genuinely impossible.
Docapybara records audio in the app and transcribes it with speaker labels. So a multilingual call comes back as a transcript where you can see who said what, in whatever language they said it in. The transcript itself stays multilingual — the system doesn't try to silently translate everything into one canonical version.
After the call, you can ask the agent to do whatever you need: "Summarize the call in English, but keep the direct quotes in their original language." Or "Pull out the three commitments the client made and translate them to English for the team." The agent works on the transcript and produces what your specific reader needs.
This means you can stop being the simultaneous interpreter. The recording does the capturing; the agent does the shape-shifting for the audience. Cross-border traders run into this constantly — see AI Notes for Import/Export and International Trade for that adjacent shape.
Drafting for the right reader, in the right register
Writing across languages isn't just about translation. It's about register — the level of formality, the conventions of the genre, the cultural defaults that aren't in any dictionary. A formal Japanese business email is built differently from a formal English one. German sales copy doesn't sound like English sales copy translated word-for-word.
The agent in Docapybara can draft for a specific reader. A typical ask might be: "Draft a follow-up email to the Japanese client based on the notes from yesterday's call. Use formal business register and include the standard closing." You give it the context, you specify the audience, you get a draft that's culturally appropriate, not just linguistically correct.
You'll still want to review and adjust — no agent gets register exactly right every time, especially for nuance — but the draft you're starting from is roughly the right shape for the reader, in the right language.
The same applies the other direction. A long, formal Mandarin document from a partner can be summarized into casual English bullet points for a quick internal sync.
A glossary that grows with your work
Anyone who works across languages eventually accumulates a personal glossary — the technical terms, the jargon specific to your field, the words your specific clients use, the translations you've worked out for tricky concepts.
A Glossary page with an inline database holds this. Columns for the term in each working language, the context where it applies, and any notes about register or alternative translations. The :::database::: directive lets the database live inside a page where you can also keep prose notes about translation choices you've made and why.
The agent can use the glossary when drafting. "Use the terms from my glossary when translating this section." It's a small thing, but it means the same English-to-French translation of a key technical term lands consistently across documents instead of drifting each time.
PDFs in any language become searchable text
A lot of cross-language work involves source documents in different languages — a Spanish-language regulation, a German technical specification, a Japanese contract draft. PDFs are the standard format and the standard headache.
Docapybara converts uploaded PDFs to markdown automatically. The conversion preserves the source language, so a Spanish PDF becomes searchable Spanish text. The agent can read it, summarize it, translate sections, or pull specific information.
A typical workflow: drop the source-language PDF on a project page. Ask the agent to "Summarize this in English, focusing on the sections about delivery timelines and penalty clauses." The summary lands on the same page. You've still got the original for accuracy, and a working English version for the team. The same workflow scales to people learning new languages — see AI Notes for Learning a New Skill.
Per-client or per-language pages, depending on your work
The right top-level structure depends on how your work splits. If you have a small number of major clients in different languages, a page per client makes sense — each client page holds the meeting notes, contracts, and correspondence in whatever language they're naturally in.
If your work splits more by language than by client — say, you do most of your work in English but handle a Spanish-speaking client base — a per-language structure can work. The agent searches across either.
Pages in Docapybara nest with no depth limit, so you can mix shapes. A Clients parent page with one child page per client; some of those children have French sub-pages and some have Mandarin sub-pages. The agent doesn't care about the structure when it searches; it reads the whole pile.
The other thing about working across languages is that thoughts don't always show up in the language you'd want to use to record them. Sometimes the right word is in your first language; sometimes a concept fits better in your second. The cognitive cost of translating-while-capturing is real and the thought often gets lost in the translation step. A new page in Docapybara is one click and you're typing. Or you tap record and talk for thirty seconds in whatever language you're thinking in. The page exists, with a timestamp. The agent can find it later when you ask in whatever language you ask in. This frees you from the translation step at capture time, which means more of your actual thinking ends up in the vault.
A starter shape that works on day one
For someone moving from scattered multilingual notes into one vault, this is roughly the shape we'd suggest. Adjust to your work.
- Inbox — uncategorized capture in whatever language. Sort later.
- Clients (or per-language top-level pages) — one child page per client or major working area.
- Glossary — terms across your working languages, with context and notes.
- Source documents — PDFs in their original language, dropped in and ready for the agent to summarize.
That's it. The vault doesn't need a translation layer; the agent does that work on demand when you have a specific reader in mind.
The thing working across languages teaches you is that fluency is partly about not having to think about which language you're in. A vault that works the same way frees up the part of your brain that used to spend energy on bookkeeping for the actual work.
Try Docapybara free — try capturing your next week of notes in whatever language each one naturally arrives in, and see how the search holds up.