One platform, grounded in your own data.

A short interview, then a working system.

Describe your plant (sites, what you make, how you make it), pick the modules you need, and answer plain-language questions about how you actually run. RetortHQ generates the entities, fields, screens, and relationships for you. Every question shows what it turns on, the fields it adds, and the rules it locks in, so nothing is a black box.

• A guided interview, not a blank schema designer
• See exactly what each answer generates before you commit
• Start from a website or an existing file and let RetortHQ propose the shape
• Load sample data to walk the whole system before you bring yours in

26 modules, 127 record types, one connected system.

Turn on only what you run. ERP covers inventory and lot traceability, formulas and BOMs, purchasing, sales, and production orders. QMS covers specifications and methods, lab results, certificates of analysis, nonconformance and CAPA, controlled documents, change control, and audits. MES covers the equipment model, ISA-88 recipes, batch execution and EBR, process data, labor, OEE, and scheduling. EAM covers the asset registry, preventive maintenance, work orders, permits and lockout/tagout, reliability, and MRO spares.

• Modules share one foundation, so a lot links to its formula, its spec, its COA, and its order
• Chemical-first: lot and sublot genealogy is built in, not an add-on
• Add a module later without rebuilding what you already have

Item master, lots, formulas, specs, COAs, orders, batches.

Every record type gets a working screen: KPI summaries at the top, search, filters, sortable columns, and a side drawer for fast entry. Marquee records get purpose-built screens, an item master with lot control and data-quality coverage, a lot detail with its inventory ledger and movements, a formula and BOM explorer, specifications, certificates of analysis, production orders, and batches.

• Consistent color and iconography: raw, intermediate, and finished read the same everywhere
• Inventory movements recorded against a lot, with a running ledger
• Role-aware: viewers read, members enter, admins delete

Trace any lot to what it came from and where it went.

Follow a finished lot back through its intermediates to the raw material lots and the suppliers behind them, or take a suspect incoming lot and see every batch and shipment it touched. This is the question a recall, a customer complaint, or an auditor actually asks.

Spreadsheets and documents into structured records.

Drop whatever you have. Spreadsheets (CSV or Excel) land in the right dataset on their own, and SDS, COA, and TDS documents (PDFs, scans, photos) are read into records using only what the document actually states. Odd column names are fine: RetortHQ reads the headers and sample values and maps them to the right fields.

• One Import all, committed in dependency order so master data lands before what depends on it
• Bad rows are held back with a plain-language fix box, while a mass update bulk-cleans the good ones
• Nothing touches live data until you commit, and Undo removes an import that went wrong

See where you stand, and what to do next.

A few plain questions place your quality process against industry best practice across specs, inspection, release, deviations, document control, and traceability. Pick your regulatory posture and the targets adjust, a general shop is not held to pharma GMP. You get a level per area and a short, ranked action plan, each item linked to the screen where you do it.

Live ecosystem data answers alongside your corpus.

Ask retrieves live ecosystem records the same way it retrieves documents, so a question about a lot, a formula, or a material is answered from your actual data and links straight back to the record. One question, both engines.

Answers with page-level citations, never a guess.

Hybrid retrieval (vector + full-text) with reranking feeds a grounded synthesis step. Every claim links to the source document and page as a citation chip. Unsupported questions get an explicit 'not in your documents' instead of a hallucination.

• Inline citations to the exact page
• Refuses to fabricate, especially on safety data
• Streamed answers

Ask across everything or narrow to specific libraries.

A multi-select control scopes a question to one or more knowledge bases, so you can ask only against the SDS library, a product line, or a department.

Structured chemical queries in the same Ask panel.

Questions like 'which products have a flash point under 100F' route to the structured chemical catalog and answer with exact values, in the one Ask experience.

Every conversation kept, collapsible, and searchable.

Conversations persist per user, with a collapsible, searchable history sidebar so past answers are easy to find and reuse.

Turn any answer into a Word, PDF, or Excel document.

Send a grounded answer straight to a formatted Word, PDF, or Excel file to share or file.

Folders, a tree, breadcrumbs, and right-click actions.

Nested folders inside knowledge bases, a persistent left tree, breadcrumb and Back/Forward/Up navigation, and a right-click menu to create, rename, delete, and move (including subfolders). It feels like Microsoft, on purpose.

• Server-side paginated and scoped: scales past 100k documents
• Search and multi-status filtering
• Bulk move and delete, move into folders

Personal, team, or organization-wide libraries.

Organize the corpus into knowledge bases scoped to yourself, a team (by member or group), or the whole organization. Access follows the same groups you manage elsewhere.

PDF, scans, handwriting, and Office files.

The pipeline extracts native-text PDFs, OCRs scans and handwriting with vision, and now parses Word, Excel (sheet-aware), PowerPoint, and OpenDocument, plus CSV, TXT, and Markdown.

• Vision OCR for scans and handwriting
• Sheet-aware Excel ingestion
• Word, PowerPoint, and OpenDocument

A durable pipeline with a visible status machine.

Each document is extracted, classified (doc type plus metadata), chunked, and embedded, with a visible status from pending to complete. Retries resume mid-pipeline and one bad file never blocks the rest.

CAS, GHS, flash point, and more, extracted and filterable.

Extracts structured chemical properties (CAS numbers, GHS classes, signal word, physical state, flash point, exposure limits, revision date) from your documents into a faceted catalog you can filter and compare.

• Faceted filters by type, hazard, signal word, and state
• Flash point and CAS number search
• Grounded in the source document, never invented

SDS currency and coverage at a glance.

A health dashboard surfaces SDS on file, aging or undated sheets, products missing an SDS, and incomplete data, so an EHS lead sees gaps before an auditor does. Also the basis of the 'SDS Health Check' as a lead magnet.

Turn a knowledge base into a grounded training course.

Generate a structured course from a knowledge base, with objectives, lessons, citations, and knowledge checks, at a chosen difficulty and depth. Every lesson is grounded in your documents.

Assign courses and programs with due dates.

Owners, admins, and group managers assign courses and programs to individuals or whole groups with due dates. Course visibility can be workspace-wide or private to assignees.

• My Courses and All Courses views
• Group managers (people managers) assign to their teams
• Due dates and 'assigned to you' surfacing

Bundle courses into an ordered path with rolled-up progress.

A program is an ordered set of courses. Progress rolls up automatically from the member courses, so a 'New Hire Safety Onboarding' path shows how far each learner has gotten.

Auto-graded exams that issue certifications.

A program's final exam is assembled from its grounded questions, unlocks once the courses are complete, and is auto-graded. Passing issues a certification, the training record an EHS manager or OSHA auditor asks for.

• Auto-graded, with a full answer review
• Certifications tab for learners
• Admin training report: assigned vs certified

An n8n-style canvas for grounded, multi-step workflows.

Build multi-step workflows on a visual canvas (input, prompt, summary, generate document, output nodes), run them over your corpus, and copy or download the result.

Describe a table, get exportable data with cited cells.

Describe the output you need ('one row per product with CAS and flash point') and Sheets builds a grounded table from your documents. Edit it conversationally, and export the exact xlsx or csv another system expects.

• Traceable cells: every AI-filled value cites its source document and page
• Spreadsheet formulas: a real calc engine (SUM, IF, VLOOKUP, and more) with cell and cross-tab references
• A formula assistant that writes a formula from a plain-language description
• Conversational editing that preserves your data (add or fill columns, add rows)
• Multi-tab workbooks, undo/redo, resizable columns, right-click insert and sort
• Import an existing xlsx or csv, then reshape with AI
• Export multi-sheet xlsx (with live formulas) or csv

Describe a purpose-built app, refine it with AI, capture data, and chart it.

Stand up a disposable app for a specific job (an R&D trial log, a QC incoming check, a stability-study tracker). Describe what you want, AI scaffolds it, and you refine it live in a builder that shows the running app as you edit. Your team then collects entries that stay in the workspace for analysis and reporting.

• Build from a document: drop a scope doc, a scanned form, a PDF, or an old spreadsheet and get a working app (spreadsheet rows import as data)
• An editable Scope tab: draft a plain-language spec, tweak it, and regenerate the app from it
• AI scaffolds and refines the app; a live-preview builder shows the running app as you edit, with drag-to-reorder
• A safe block model (form, table, stat, chart) over typed datasets, no code
• Computed fields with a formula engine, barcode/QR scan fields (device camera), and catalog reference fields that auto-fill CAS, hazard, and flash point from your chemical catalog
• Charts (bar, line, scatter, pie) and stats that aggregate your entries, with an optional linear-regression trendline and SPC control charts
• Share an app workspace-wide or keep it private, and assign people or groups as users or editors
• Captured data is structured and stays for later analysis and reporting
• Captured entries flow into Ask: ask questions over your app data and get answers cited back to the app
• Publish an app for the team, or keep it a draft while you build

Drop data or point at an app, get a dashboard.

Build a KPI dashboard over a spreadsheet you drop in or a live app dataset, optionally grounded in your knowledge bases. Headline numbers, charts, and tables, laid out automatically, with an interactive builder popup and export to JSON, CSV, or print/PDF.

• Pull from a live app dataset or drop a spreadsheet snapshot, and ground the design in knowledge bases
• Auto-laid-out KPIs, charts, and tables from a plain-language prompt
• Export a dashboard as JSON, its data as CSV, or print to PDF
• Private or workspace-shared dashboards

Produce Word, PDF, and Excel outputs.

Generate formatted Word, PDF, and Excel documents from answers, agents, and sheets, ready to hand off.

A quick, self-descriptive type and strengths.

Each member can take a short, opt-in assessment: a four-axis type in the Myers-Briggs tradition plus a ranked strengths inventory. Self-descriptive, never diagnostic, and the member controls who can see it.

• Opt-in and consent-tracked; visibility is self, managers, or the whole workspace
• Type, strengths, and temperament shown on the profile, with retake and clear

Assign a role lens like QC, Ops, or Plant Manager.

Admins define role archetypes (a job description plus tone, depth, and priorities) and assign one to each member. A starter set for chemical manufacturing is one click away.

• Manufacturing starter set: Plant Manager, QC/QA, Operations, R&D, MRP, Customer Service, Executive
• Create and edit your own role lenses

The same cited answer, framed for who is asking.

Ask frames its answer for the reader's persona and working style: an executive gets the bottom line, a QC analyst gets methods and spec limits. Facts, sources, and the not-in-corpus guardrail never change. On by default, with a per-user and a workspace-wide off switch.

Import documents from Google Drive into a knowledge base.

Connect Google Drive and browse-and-import files straight into a knowledge base, where they process like any upload.

Owners, admins, members, viewers, and AD-style groups.

First-party tenancy with an Organization hub for people, groups, and roles. Groups (R&D, QC, EHS) drive knowledge base access and training assignment, and group managers act as people managers.

No workspace can ever see another's data.

Isolation is enforced at the database with row-level security, backed by explicit tenant filters and a cross-tenant test suite. It is the product's first promise.

Simple per-seat plans, no platform fee.

Stripe-powered per-seat plans with Checkout and a billing portal, plus AI credit metering and a monthly budget so costs are predictable. No platform fee, ever.

Dark and light, teal brand, molecular accents.

A consistent design system with a teal brand, glass chrome, subtle molecular 3D accents, toasts, and reduced-motion support across the whole app.