New in the playground
Dump it in any language. Ask in yours.
Drop in files, Google Drive, Google Docs and Outlook. Then ask your memory a question in any of 15 languages — the answer comes back in that language, built from every source, whatever it was written in.
Try it below — pick a language, hover a row to see which source it came from.
q3-launch · OctaMem (default model)
Ask in English· waiting for the dump to finish
Dumping sources into memory…
German mail, an English doc, a Japanese PDF.
Ask anything about your memory…
Replies use this memory’s context, in the language you ask in.
01One memory, no language wall
Written in three languages. Remembered as one.
Your team doesn’t pick one language, so your memory shouldn’t either. Each source is parsed into facts — who decided what, and when — and the language it arrived in becomes provenance, not a barrier.
Multilingual AI memory means one store of company knowledge you can question in any supported language, regardless of the language each source was written in. No per-language indexes, no translation step to manage, and no setting to change between questions.
01Dumped in any language
OutlookDE
Wir verschieben den Launch auf den 18. September – bitte Kalender anpassen.
Google DocsEN
Final call: €49 per seat, billed annually. No free tier at launch.
Google DriveJA
ベータ版は200チームに限定し、9月以降に段階的に拡大します。
02Stored as meaning
q3-launch memory
- launch_date2026-09-18
- price_per_seat€49 · annual
- beta_cap200 teams
Facts, owners and sources are kept — the original wording and its language are just provenance.
03Answered in yours
Decisions
- Launch moved to September 18
- Price set at €49 per seat
- Beta capped at 200 teams
0215 languages
Fifteen ways to ask. One memory answering.
Hover a language to read the question in it. Click to ask the playground above.
03Memory dump
Files, Drive, Docs, Outlook. And the rest.
Upload a file or connect a source once from the Memory dump screen. Everything you add becomes typed, source-linked memory, ready to ask in any language.
File upload
PDF, DOCX, PPTX, XLSX, CSV, EML, TXT, JSON — dropped straight into a memory.
Google Drive
Folders and shared drives, re-synced as files change.
Google Docs
Docs read as structured text, headings and tables kept.
Microsoft Outlook
Mail threads with sender, recipients and timeline intact.
OneDrive
Personal and work files from your Microsoft 365 tenant.
SharePoint
Team sites and document libraries, permissions respected.
Microsoft Teams
Channel conversations and the decisions made in them.
OneNote
Notebooks, sections and pages as searchable memory.
04Who it’s for
Built for teams that don’t share a language.
Global teams
Headquarters writes in English, the Berlin office emails in German, the Tokyo team ships specs in Japanese. Everyone asks in their own language and gets the same facts.
Cross-border legal work
Contracts, correspondence and filings in several languages sit in one matter memory. Ask what was agreed and see the clause it came from.
International real estate
Investor updates, leases and broker emails from different markets answer one question: what changed on this asset, and who said so.
Support and success
Tickets and threads arrive in the customer’s language. Agents and AI assistants answer from all of them, in whatever language the next customer writes.
05Over MCP
Same memory. Every agent, every language.
The OctaMem MCP server reads the same memory as the playground. Ask from Claude, Cursor or Codex in Hindi, Arabic or Korean and the results come back in that language, sources attached.
› Q3 लॉन्च के लिए हमने क्या तय किया?
⏺ octamem · search_context
⎿ { "answer_language": "hi",
"results": [
{ "fact": "लॉन्च 18 सितंबर तक खिसकाया गया",
"source": "Outlook", "original_language": "de" },
{ "fact": "कीमत €49 प्रति सीट तय",
"source": "Google Docs", "original_language": "en" },
{ "fact": "बीटा 200 टीमों तक सीमित",
"source": "Google Drive", "original_language": "ja" }
] }⏺ Q3 लॉन्च के लिए तीन फ़ैसले लिए गए, जो आपके तीन स्रोतों से मिले।
06Questions
Multilingual memory, answered.
Do my sources need to be in the same language as my question?
No. A German email, an English doc and a Japanese PDF can sit in the same memory. Ask in any of the fifteen languages and the answer draws on all of them.
Is anything translated when I dump a source?
Your originals are kept as they were written. What OctaMem stores is the meaning — the facts, owners and dates — with the original passage attached as the source, so you can always open the exact wording.
Does the MCP server answer in my language too?
Yes. The MCP server reads the same memory as the playground. Ask from Claude, Cursor, Codex or any MCP client in Hindi or Korean and the results come back in Hindi or Korean.
What about right-to-left languages?
Arabic is fully supported, including right-to-left layout for answers and tables in the playground.
Which languages can I ask in?
English, Spanish, French, German, Japanese, Chinese (Mandarin), Greek, Hindi, Arabic, Portuguese, Russian, Korean, Italian, Turkish and Dutch.
Do I have to set a language anywhere?
No. Ask in the language you want the answer in. The next question can be in a different language, against the same memory.
Which files and sources can I dump into memory?
Upload PDF, DOCX, PPTX, XLSX, CSV, EML, TXT and JSON files, or connect Google Drive, Google Docs, Microsoft Outlook, OneDrive, SharePoint, Teams and OneNote from the Memory dump screen.
Can I see where an answer came from?
Yes. Every answer links back to its sources. Open the memory response under an answer to read the original passages, in the language they were written in.
Organisational intelligence compounds
Two minutes to set up. Smarter every day after.
Connect your sources once. Every file, commit, email, and decision after that feeds the memory, and everyone who asks it gets sharper answers.
Builder includes 5,000 MOps and 10 GB for $49/month.