What can I answer for you?
The agentic document workspace. Ask, and every claim comes back with a citation.
See what it readsReads the files you already have
Drop in a folder of contracts, a spreadsheet, a deck or a photo of a whiteboard. Scans and pictures are read with OCR, and every answer points back to the page, cell or slide it came from.
Read with OCR where it is marked. How OCR works
- PDFOCR
- Page by page; a scanned page is read by OCR and cited by its page number.
- Word and OpenDocumentOCR
- The whole text; a Word file that is only pictures of pages is read by OCR.
- .docx.odt.rtf
- Spreadsheets
- Sheet by sheet, cited by the cells an answer came from.
- .xlsx.ods.csv.tsv
- Presentations
- Slide by slide, without the speaker notes.
- .pptx.odp
- Photos and scansOCR
- The words in the picture: a receipt, a whiteboard, a screenshot of a clause.
- .png.jpg.webp.avif.tiff.bmp.gif
- Web pages
- By its link or as a saved file, kept as the article without the menus around it.
- .html.htm
- A message and its parts, with who sent it and when.
- .eml
- Transcripts
- Captions with their times, so an answer can point to the moment.
- .vtt
- Text
- As written, Markdown headings kept.
- .txt.md
- Data
- Each value beside its name, so a figure is found by what it is called.
- .json.xml
Watch a document become answers
From the moment the bytes leave your computer, the row says what is happening: uploading, each page read, each passage indexed, then ready. No spinner, no bar parked at seventy percent.
What notice does termination need?
Either party may end the agreement with ninety days’ written noticep. 7
Illustration. In the workspace, every number on this bar is measured as the file is read.
How it works
Add your documents
Drop files or a whole folder into a project, or paste a link to a web page. Each one is read, split into passages and indexed for search.
Ask, and every claim is cited
Ask in plain words. The answer cites the passages it stands on, and a sentence with nothing behind it is marked as such rather than dressed up.
Make something from it
Turn the answer into a note, a sheet or a deck in the same project. Sources travel with what you make, and every format exports.
Three makers, one conversation
Ask in the chat; when what you want is a document, the right maker builds it from the same sources, in its own tab.
- NoteAI Researcher
Plans a report section by section, reads your documents and the web, and writes it with every claim cited.
How the AI Researcher works - SheetAI Analyst
Builds a table from your documents and works out figures over every row, each cell carrying its source.
How the AI Analyst works - SlidesAI Designer
Turns an answer or a brief into a deck you can edit, present, and export to PowerPoint or PDF.
How the AI Designer works
Every claim opens its page
Click a citation and the source opens at the page, the cells or the slide it came from, with the passage marked. You check the claim, not the summary of it.
How citations workPersonal details leave first
Email addresses, phone numbers, card numbers, national identifiers and people's names are removed from a passage before it reaches a model.
Your workspace is walled in the database
Every row belongs to one workspace, enforced by the database itself, so another workspace cannot read your documents even by mistake.
Not used for training
SYNTHIEN does not train models on your documents. A passage is sent to a model only to answer the question you asked.
- The whole of it is in the privacy notice.
Pay for the questions you ask
Buy credits once and spend them as you ask. There is no subscription and no seat fee; reading and indexing your documents costs nothing.
- Starter$10about 675 everyday questions
- Plus$25about 2,027 everyday questions
- Pro$60about 5,405 everyday questions
Questions people ask first
What files can it read?
PDF, Word, Excel and PowerPoint; their OpenDocument twins; CSV, TSV, RTF, text, Markdown, JSON and XML; web pages, email and subtitle files; and pictures — PNG, JPEG, WebP, AVIF, TIFF, BMP and GIF. A file can be up to 64 MB.
Does it read scanned documents?
Yes. A PDF page with no text layer, a Word file that is only pictures of pages, and a photo or screenshot are read with OCR, and the document's row says how many pages OCR read. You watch each page being read as it happens.
Where are my documents kept?
In the workspace you upload them to: the service's own database and file storage, on servers in Google Cloud's Singapore region, reached through Cloudflare. Another workspace cannot read them: every row is fenced to its workspace in the database itself, not only in the application.
Are my documents used to train models?
SYNTHIEN does not train models on your documents. A passage is sent to a model only to answer your question, and email addresses, phone numbers, card numbers, national identifiers and people's names are removed from it first. The default models are OpenAI's, called through OpenAI's API, whose terms say data sent through it is not used to train their models. If you pick a DeepSeek model, the passage goes to DeepSeek under its own terms instead. The Privacy page has the detail.
What does a question cost?
You buy credits once and spend them as you ask; there is no subscription. An everyday question costs one or two cents, a harder one under ten, and the flagship models more. Pricing lists every tier.
Can I take my work out?
Yes. Answers export to Markdown, Word, PDF, BibTeX and RIS; notes to Markdown, Word, PDF and HTML; sheets to Excel, CSV and Markdown; decks to PowerPoint, PDF and handouts. Citations travel with them.
Can my team use it together?
Yes. A workspace has members: owners invite people and choose what each may do, and a viewer can read without changing anything. A chat, its notes and its decks can be shared with the people in it.
How do I get help?
The guide explains every part of the product. From the dashboard, Support searches the guide, files a report you can follow by its number, and shows whether any part of the service is down.