Ask your documents.
Get answers you can check.
PageFinder reads the contracts, invoices and reports you already keep in folders and cloud storage, answers questions about them with page-level citations, and turns what it finds into Excel. Your originals never move.
Get started →- Indexed in place — never copied
- Every answer cites its pages
- Per-workspace access control
Your documents already have the answers
The answer is in a PDF somewhere — but which one?
Ask in plain language and get the answer with the document and page it came from, instead of opening files one by one.
Search finds the word, not the meaning
Semantic and keyword search run together: “termination for convenience” finds the clause however it is worded, and “INV-20418” still finds the exact invoice.
Scans and stamped forms are invisible to search
Pages without a text layer are read as pages, so scanned invoices, signed forms and tables are searchable like everything else.
Moving documents into yet another system is a project
Point a workspace at the bucket or drive where the files already live. Nothing is uploaded, moved or reorganised.
How it works
Connect a source
Point a workspace at an S3 bucket, Google Drive, OneDrive or Box. PageFinder indexes the files where they are — nothing is uploaded or moved.
Ask in plain language
Ask it the way you would ask a colleague: “What is ACME’s payment term?” The agent searches, opens the right pages and reads them.
Check it, then use it
Every answer links to the pages it came from. Ask for a table and get an Excel workbook back.
Built for documents that matter
Originals stay put
Documents are read from their source when needed. Only the search index and the files PageFinder generates are stored — there is no second copy to secure.
Cited, not guessed
Answers point to the document and page they came from, so you can open the original and verify in a click.
Reads scans and tables
Pages without a text layer are looked at, not skipped. Invoices, stamped forms and charts are understood as pages, not just as text.
Search by meaning and by word
Semantic and keyword search run together, so “termination for convenience” finds the clause and “INV-20418” finds the invoice.
Workspaces with real boundaries
Each workspace is its own hard partition with its own members and roles. Members see the documents they were given access to, and nothing else.
Answers that become spreadsheets
“Collect the amounts and due dates from the ACME invoices” ends in an Excel file, with the audit trail of how each row was found.
Your documents stay yours
- No second copy. Originals are read from your storage on demand. What is kept is the search index, page previews and the files generated for you.
- Encrypted. HTTPS for every request, and everything stored — databases, the search index, generated files and the access tokens for your storage — encrypted at rest.
- Not used for training. AI models run through Amazon Bedrock by default, under terms that keep your content out of model training.
- Isolated workspaces. Access is enforced by the server on every request, workspace by workspace.
- Read-only access to your sources. PageFinder asks for read-only access where the provider offers it, and never edits, moves or deletes your files.
- Private deployment available. For organisations that need it, the whole system can run inside your own environment.
Common questions
Are my documents copied or uploaded?
No. PageFinder indexes files in place and reads the original from your storage only when it needs a page. What it keeps is the search index and the files it generates for you, such as workbooks.
Who can see what?
Only members of a workspace can ask about its documents, and a member’s role decides what else they can do. Invite colleagues by email from your account.
What kinds of files does it read?
PDFs — digital and scanned — Word, Excel and PowerPoint files, email messages and images.
Is my content used to train AI models?
No. Models run through Amazon Bedrock by default — or Google Gemini, if your workspace administrator chooses it — under terms that do not allow training on your content. The Privacy Policy names each provider.
Which storage can I connect?
Amazon S3 buckets — ours or your own — Google Drive, Microsoft OneDrive and Box.
What does it cost?
You pay for what you use. Add prepaid credit by card and each indexed file and answer draws it down at the prices shown in your account; new accounts start with a credit to try it. A prepaid plan may also carry a small monthly maintenance fee based on how many files you keep indexed; your account shows it before it applies. Turn on automatic top-up if you would rather not watch the balance.
How do I get an account?
Sign in with Google, or with an email address and password. If your address is not yet allowed on this deployment you will be told so, and you can ask its administrator to invite you.
I can’t reset my password.
If you signed up with Google, your account has no PageFinder password to reset — choose “Continue with Google” on the sign-in page instead. An address can use Google or a password, not both.