Extract data from safety data sheets
Unstract reads supplier safety data sheets, from 16-section GHS documents to older MSDS layouts in several languages, and sends the product, hazards, composition and revision date to your EHS or chemical inventory system.
H317 is new vs rev 3 — relabel + update inventory, RT
SDS NG-0418 · VERSION 4.0 · SUPERSEDES 03/02/2024
{ "product": "Corvane HD-40", "product_code": "CV-HD40-20L", "supplier": "Northgate Chemical Supply", "revision_date": "09/18/2026", "signal_word": "Danger", "hazard_statements": "H225, H315, H317, H319", "main_component_cas": "5989-27-5", "flash_point_c": 18, "un_number": "UN1993", "transport_class": "3", "confidence": 0.94}
Trusted by manufacturing and quality teams around the globe
Every supplier writes its safety data sheet a little differently
A current 16-section SDS sent as a clean PDF with the first order is the easy case. The chemical binder at most plants holds a lot more than that.
An MSDS from the plant binder, photocopied for years before anyone scanned it
One product, three formats still on file
A binder copy marked up by hand with the storage area and the drums on hand
The supplier's SDS in four languages, sent as one PDF
A distributor's product list with a revision date per line
A compliance specialist opens each new sheet, compares it with the copy on file and retypes the changes, so a revised hazard can sit unnoticed while the drum is already in use on the floor.
Unstract takes on the reading — it pulls the product, supplier, hazard statements, composition and revision date from each sheet in the supplier's own layout, compares them with the version on file and sends the record to your EHS or chemical inventory system.
Read, extract and keep your SDS library current on one platform
Parse the sheet, turn its sections into fields, then run the revision check around them. Each product works alone, and the three together keep the library in step with what suppliers ship.
Parse each safety data sheet with every section kept in order
Capture the full text of a 16-section GHS safety data sheet, an older nine-section MSDS or a scan from the plant binder, with the layout preserved. Section numbers stay attached to their content, so a flash point from Section 9 never lands in the transport block, and a composition table keeps each CAS number on the line of its chemical.
Sheets in Spanish, French or German come through as text in their own language. Pass it to your own LLM, or make parsing the first step of an SDS workflow.
Learn more about LLMWhispererA safety data sheet data extractor that compares each revision
Name the fields your EHS team keeps for every product, and each sheet comes back in the same JSON structure, whether the supplier calls it a safety data sheet, a material safety data sheet or a Sicherheitsdatenblatt. The new revision is set beside the one on file as it comes out: a hazard statement that wasn't there before, a changed composition range or a sheet older than your review cycle arrives flagged with the reason.
Call it as an API when a product is set up, run it as an ETL pipeline over a supplier inbox or a folder of scanned binders, or make it one step in an orchestrated workflow.
Learn more about Document StudioKeep every safety data sheet on file at its latest revision
Describe how your site handles a new or revised safety data sheet in plain English, and Flow's agents build the workflow. Sheets are picked up from supplier emails or a shared folder, each one is read, and the product, hazards, composition and revision date are compared with the version on file. Records go to your SDS library or EHS system. You approve the workflow once, and every new sheet follows it.
Set the points where a person reviews: a new hazard class, a changed signal word, a composition range that moved, a sheet past the age your program allows. Only those products wait, with both versions side by side. Unstract reads what the supplier wrote; it does not classify hazards or give safety advice, and your EHS team decides what changes on the floor.
Learn more about FlowRead your supplier SDSs today. Add the revision check when you're ready.
Start for free to parse and extract your own safety data sheets now. Setting up the revision workflow starts with a consultation.
Safety data sheets in every format a supplier still sends
Documents
- Safety data sheets (GHS, 16 sections)
- Material safety data sheets (older formats)
- ANSI Z400.1 safety data sheets
- Revised supplier SDSs
- Multilingual supplier SDSs
- Extended SDSs with exposure scenarios
- Technical data sheets
- GHS container labels
- Certificates of analysis
- Regulatory compliance statements
- Dangerous goods declarations
- SDS revision notices from suppliers
- Distributor product lists with revision dates
- Site chemical inventory lists
Fields captured
For the people who answer for every chemical on site
Every product on site matched to a current sheet, with the ones past their review age listed before an inspector asks for them.
Hazard statements, CAS numbers and composition ranges in one record per product, ready for regulatory reporting and substance checks.
The signal word, pictograms and first-aid details for a product at hand when a new drum reaches the line, taken from the supplier's latest sheet.
Each manufacturer's sheet read on receipt and its revision date logged, so customers get the current version with every shipment.
A chemical inventory that keeps up with supplier revisions
Formulation details and site inventories stay private
Section 3 of a safety data sheet gives a supplier's composition ranges, the closest a customer gets to the formulation, and suppliers share them under that trust. Your chemical inventory shows which substances each plant stores and in what quantity.
- Encrypted in transit and at rest.
- Supplier sheets and inventory data are never used for model training.
- Access by role, site and supplier, with every correction and review decision logged against the product.
- You set retention, so superseded sheets stay as long as your records program requires and are then deleted.
- Self-hosted deployment for manufacturers that keep chemical data inside their own network.
What EHS and chemical compliance teams ask first
Route supplier sheets to Unstract from an inbox, a shared folder or an API call. LLMWhisperer reads each sheet with its sections in order, and Document Studio returns the product, supplier, hazards, composition and revision date as JSON for your EHS or chemical inventory system. Safety data sheet extraction runs the same way on the first sheet and the five-hundredth.
Yes. Older material safety data sheets in the nine-section OSHA layout or the ANSI Z400.1 format are read alongside current 16-section GHS sheets. Document Studio finds each field by meaning, so a flash point printed under "Fire and explosion data" on an old MSDS comes back in the same field as one from Section 9.
Yes. LLMWhisperer reads photocopied and scanned sheets, including pages that have been copied many times, and keeps each CAS number on the row of its chemical. Handwritten notes added on site are read as notes and kept apart from the supplier's text.
Yes. Sheets in Spanish, French, German and other languages are parsed in their own language, and Document Studio returns the fields you named in your schema. H-codes, CAS numbers and UN numbers stay the same in every language, which makes them a reliable anchor.
Each sheet's product code and revision date are compared with the record on file. A newer revision updates the record, and a product whose sheet is older than your review cycle is listed, so your team knows which suppliers to ask for a current version.
The new sheet is compared field by field with the previous one. An added hazard statement, a different signal word or a composition range that moved arrives flagged, with the old and new values side by side and the revision dates of both sheets.
No. Unstract reads what the supplier's sheet states and reports it. It does not classify chemicals, author or correct an SDS, or give safety advice. Your EHS team decides on labels, storage and training.
Both, in layers. LLMWhisperer is the safety data sheet parser, Document Studio returns your fields, and Flow runs the safety data sheet processing around them: intake, the revision check, review and the update to your SDS library.
Yes. Document Studio deploys an endpoint for the document type: POST an SDS and get the product, hazards, composition and transport details back as JSON. The same extraction runs as an ETL pipeline over a folder or bucket of sheets.
Unstract is SOC 2, ISO 27001, GDPR and HIPAA compliant. Sheets and inventory data are encrypted in transit and at rest, never used to train models, and kept only as long as you set. Sites that keep chemical data on their own network can self-host; the Trust Center has the full list of controls.
A mix of supplier sheets, current and old, scanned and digital, plus the fields your inventory keeps for each product. We set up the extraction and the revision check around them in a consultation. To try it on your own, start for free and upload one sheet.
Bring your chemical binder to the consultation
GHS sheets, older MSDSs and supplier sheets in other languages, with the inventory they belong to. We read them with you and show which products have a newer revision or a hazard your records miss.