Ask questions of your data

The RAG agent finds answers, fast. It pieces them together from scattered data. You only pay for what you use.

Add your first document
  • PDF
  • Word
  • PowerPoint
  • Excel
  • Images
  • Text
Add your first document…

€5.00 free: 500 pages of documents and 25 questions · no credit card

And chat about it from your messaging appsSlackTelegramMicrosoft Teams
  • Correct answers88.1% · FinanceBench
  • Indexing€5 / 1,000 pages · €2.50 in batch
  • To get started€5.00 free: 500 pages of documents and 25 questions
  • Citationsopen at the page
  • Datahosted in the EU
  • FormatsPDF · DOCX · PPTX · XLSX · images
Data sources

Connect your data wherever it lives.

Connect them from your workspace in a few clicks or upload your files: everything goes into one index, queried from the chat or your messaging apps.

Agentic Searchone index

  • Connectors

    • Notion: Pages and databases
    • Google Drive: Docs, Sheets, Slides and PDFs
    • SharePoint and OneDrive: SharePoint sites and OneDrive files
    • Confluence: Wiki spaces and pages
    • GitHub: READMEs, documentation and issues
    • Dropbox: Files and foldersSoon
    • Outlook: Microsoft 365 emailsSoon
    • Zoom: Meeting transcriptsSoon
    • Airtable: Base tablesSoon
    • Asana: Projects, tasks and commentsSoon
    • Trello: Boards, cards and commentsSoon
    • Google Calendar: Calendar eventsSoon
    • Outlook Calendar: Calendar eventsSoon
    • Mailchimp: Sent campaignsSoon
    • Calendly: Meetings and inviteesSoon
  • Files

    • PDF: .pdf
    • Word: .docx .doc .odt .rtf
    • PowerPoint: .pptx .ppt .odp
    • Excel: .xlsx
    • CSV: .csv
    • Text: .txt .md
    • HTML: .html .htm
    • Images: .png .jpg .webp
  • 5 connectors (10 more coming soon) and 19 file extensions, brought together in one index.

    Connect a source
Agentic SearchRAG engine

Your RAG data is indexed in Europe

Then chat

  • Web chat
  • Slack
  • Telegram
  • Microsoft Teams
Cited sources

Every claim can be checked.

The answer cites its passages: hover a number to read the excerpt, click to open the document at the cited page, with the passage highlighted.

AMD_2022_10K.pdfannual report · 121 pages

Does AMD have a reasonably healthy liquidity profile based on its quick ratio for FY22?

Source relevance 92%

Yes. AMD's quick ratio at December 31, 2022 is 1.57: its most liquid assets cover its short-term liabilities more than one and a half times.

  • Cash and cash equivalents: $4,835M 1
  • Short-term investments: $1,020M 1
  • Accounts receivable, net: $4,126M 1
  • Total current liabilities: $6,369M 2

(4,835 + 1,020 + 4,126) / 6,369 = 1.57. Inventories ($3,771M) are left out, as they don't turn into cash right away 1.

Cited passages
Accuracy

This RAG and how accurate it is

We evaluated this RAG on FinanceBench, one of the toughest benchmarks for accuracy on reports full of tables and numbers: the system has to find the right answers across 12,000 pages, 84 documents and 150 questions.

  • 88.1%

    correct answers

    119 of 135 questions judged correct

  • 84

    SEC filings

    Quarterly results from AMD, Boeing, Pepsi and more

  • 12,000

    pages indexed together

    finding the right report among all those pages

  • 135

    analyst questions

    ratios, trends, facts, yes / no

  • ~11 s

    per answer

    search, verification and citations included

A classic RAG compared with Agentic Search

Our RAG works with navigation tools: the model itself goes looking for the answer, weighs it, goes back, and reads the context around key passages.

Without tools

classic RAG: a single pass

Baseline

46.0%correct answers

Measured on the full corpus: 84 reports, 135 questions

  • 64 correct
  • 20 wrong
  • 55 refused

Receives the 10 best passages and answers in a single pass. If the evidence isn't there, it can't go and get it.

  • A table split across two passages stays unreadable
  • A footnote on page 87 is never read if the search missed it
  • Without evidence, it refuses or guesses
Wrong answers or refusals
54.0%
Generation
~6.7 s
Tokens used
~6,000

With our tools

the agent goes and reads the document

Agentic Search

88.1%correct answers

Measured on the full corpus: 84 reports, 135 questions

  • 119 correct
  • 15 wrong
  • 1 refused

Starts from the same passages, then looks for what's missing: searches again, opens a document, reads the page, checks the surrounding context, reads the whole page with our navigation tooling search grep read_page open_doc navigate .

  • Reads the full table on the page, headers included
  • grep confirms whether a term appears anywhere in the document
  • Every number is cited, the page opens at the passage
Wrong answers or refusals
11.9%
Generation
~11.0 s
Tokens used
~18,000
Pricing

Pricing

Free to start, then you only pay for what you use, at the end of the month.

Free

Try it on your own documents, no credit card needed.

€0no card

  • €5.00 free: 500 pages of documents and 25 questions
  • Up to 500 pages of documents, any number of files
No commitment

Pay as you go

You only pay for what you use, billed at the end of the month.

€0a month, then pay as you go · tax incl.

  • Question: €0.10, cited and verified answer
  • Indexing: €5 per 1,000 pages, €2.50 in batch (2 or more documents uploaded together)
  • Example: ≈ €59.00 for 15,000 indexed pages and 215 questions
  • Up to 2,000 documents, 20,000 pages
  • Your free credits are used first
  • Charged to your card on the 1st of the month, invoice by email, cancel anytime

What's included

FreePay as you go
Indexed pages500 in totalup to 20,000
Indexing€0.75 for 150 pages€5 per 1,000 pages · €2.50 in batch (2 or more documents uploaded together)
Questions€5.00 free: 500 pages of documents and 25 questions€0.10 per question
FormatsPDF, Word, PowerPoint, Excel, images, textSame
CitationsOpen in the document, at the pageSame
Storage and browsingFreeFree
DataHosted in the EU, isolated per accountSame
API and MCPIncludedIncluded
PaymentNo cardCard, charged on the 1st for actual usage

How much does 1 GB of documents cost?

Indexing is priced per page, not per byte. Across FinanceBench's 84 reports, 165 MB make 12,013 pages, about 14 KB per page of text: 1 GB of text PDFs is roughly 71,000 pages. An illustrated or scanned document weighs far more per page, so it costs less per GB.

1 GB of…Size per pagePagesIndexing
Text PDFsannual reports, contracts, memos≈ 14 KB≈ 71,000≈ €177.50in batch
Illustrated PDFsslide decks, brochures, charts≈ 100 KB≈ 10,000≈ €25.00in batch
Scanned documentsletters, scanned invoices≈ 250 KB≈ 4,000≈ €10.00in batch

You pay for indexing once; after that, each question costs €0.10, however much you've indexed. Beyond 20,000 pages, let's talk: Team plan on quote.