Sheaf.us
Pricing Login

Give Claude Document Superpowers.

Keep your Data. Keep your Agents. Keep your Edge.

How is SpaceX’s AI segment doing?

AI revenue was $2,561 million in the second quarter of 2026.43 The segment still lost $1,257 million from operations in that quarter.44

Q2 2026 earnings releasepage 3 · 2 citations

See citations on the real document.

Engineered for:

Cadence Bank Old National Bank Webster Bank Hanover Insurance Cincinnati Financial Stewart Title
COMPARE

What Sheaf does. What others do.

JobSheafThe industry
IntakeBeautiful, branded intake forms. Request links chase the missing document for you.An upload portal, then an email thread for whatever is missing.
ConvertConvert, split, classify and extract, per docType. Optimized for analysis.A dump of OCR text, and templates someone maintains by hand.
StoreSecure, agent-optimized structure. Single-click migration in and out; point any agent at it.Locked in a vendor database. Export means a CSV and a ticket.
SearchUltra-fast search structure built for intelligence and research across every project.Keyword search, one folder at a time.
CiteForensic citation of every word or value on each page, with a direct-link preview.A filename, if you are lucky.
Your agentUse your own agent to open files, ask questions and manage projects.Their chatbot, their model, their browser tab.
ModelsModel-agnostic, with built-in evals and scoreboards. Compare models to save & improve accuracy.One model, picked by the vendor. No way to test another.
TeachCustom intelligence, taught by your team: accept or reject, add tests, give directions.A model trained on someone else’s documents.
AutomateAutomates processing and customer feedback on KYC, onboarding and applications.A queue that someone works through on Monday.
WhitelabelSelf-hosted or managed whitelabel, under your own brand. Your data stays in-house.Their brand on every screen. Your data on their servers.
RESEARCH

Verify citations just by hovering on the sentence.

Get context by just clicking on the thumbnail and the full page opens.

Drop your hardest document here.

PDF, PNG or JPEG · up to 100 MB

Or inspect a finished case

Automate your document workflow. Simply.

05 / YOUR DATA

You keep your data + pain‑free migration.

  1. Your entire corpus of documents can be stored on your own S3 bucket (which you can set up in minutes). This means the data remains yours.
  2. A standard, agent‑friendly format of doc storage and folder structure means you can copy or migrate a project, or even your entire document library, seamlessly.
S3 bucketIn‑house Sheaf
S3 bucketAWS, customer‑managed
Sheaf Intelligent Document Management
Intake forms
Human‑in‑the‑Loop review
Classification & processing steps
Versatile workflows
Forensic fingerprinting & extraction
Research with pixel‑level citation
01 / INTAKE

Auto‑split complex, merged documents.

A merged, scanned PDF becomes 43 documents; named, classified and extracted perfectly.

2026-04-30_JANE-DOE.PDF 412 pages · 43 documents found
  • pp. 1–6URLA — Form 1003required
  • pp. 7–9Lender loan informationrequired
  • pp. 10–12Credit report — soft pullrequired
  • pp. 13–28Bank statements — 3 accountsrequired
  • pp. 29–31Unlabelled scanreads as title commitment
  • —Certificate of insurancenot in this packet
02 / CRITERIA

Make an “Expected Doc List” and say what you need from each.

All intelligence starts from your foundation. The “Setup” agent configures your lists and extract schema per doc in seconds.

Expected documents Purchase loan · California

  • URLA — Form 1003signed
  • Credit reportunder 90 days
  • Bank statements3 months
  • Certificate of insurancecurrent term
  • Government IDunexpired

Schema Bank statement

  • account_holderstring
  • period_startdate
  • period_enddate
  • ending_balancecurrency
  • nsf_countinteger

Change the list and the whole corpus re‑scores against it. Nothing is read twice.

03 / GROUNDING

Fingerprint (parse with coordinates per element) to the exact pixel.

Every value carries its page and a four‑number box — $9,502.93, page 3 of 6. Any number drills straight back to the pixel it was read from.

What comes back for one field

label
Borrower — monthly income
value
$9,502.93
page
3 of 6
box
0.612, 0.284, 0.788, 0.301
grounding
pdf‑text‑layer
Technical detail Why a value has to point at a page, and what an agent gets when it does Read the spec →
04 / REVIEW

Human‑in‑the‑loop review: building custom intelligence.

Three decisions per document — page span, requirement match, field values — each confirmed or redirected from the keyboard. Redirects are the training signal.

The Sheaf review room. A packet is open at URLA Form 1003, pages 1 to 6 of 43 documents, with page thumbnails across the middle, a filing panel on the right with its three decisions on the document, and a list of six page decisions each marked inherited good. Every stage on one screen The scan itself, not a summary Match and file, or send it back A decision on every page 43 documents, re‑cuttable
Page decisions read inherited good where an earlier human decision already covered those exact pages. Matching is what files the document. Each decision is bound to a page span, so re‑cutting the boundaries around it changes nothing.
TRY IT

1‑click to see an analysed doc & case.

Then upload your own, and see how it is processed.

Single‑page form

Vehicle accident report

One page. Every field read, and every value pinned to the place on the page it was read from.

One PDF · 35 pages · 6 documents

A merged packet, split for you

  1. Claim noticepp. 1–3
  2. Police reportpp. 4–9
  3. Repair estimatepp. 10–15
  4. Medical recordspp. 16–27
  5. Invoicespp. 28–32
  6. Correspondencepp. 33–35

Where each document starts and ends is suggested for you. Accept a split, move it, or merge two back together.

What document work actually requires

Page-level citations

Every extracted value carries the file, page and span it came from. An answer you cannot walk back to the page is an answer you cannot defend.

Scanned and photographed documents

Real work arrives as a scan of a print of a fax. A page with no text layer is the ordinary case here, not the exception that gets routed to a person.

Knowing what is missing

Completeness is the question most systems never answer. Sheaf names the requirement that nothing satisfied, which is usually what is holding the file up.

Audit trail and reviewer decisions

Every automatic outcome points back at the human decisions behind it, so an auditor can walk from a value to a page without leaving the record.

Security and data handling

Every request is checked against the role that made it, records age out on a fixed retention window, and each value stays traceable to its page.

Sheaf / Test Environment

Drop your files here

or click to choose

Multiple PDF, PNG, JPEG & TIFF accepted

Selecting files opens your OS picker. Processing continues in Sheaf. No files are uploaded from this page.

Compare
Scoreboards