Freight forwarding & logistics
Shipment documents, customs paperwork, charges, parties, and delivery evidence.
Explore workflowIntelligent document processing (IDP)
Precision document processing for teams that can't afford mistakes.
Extract nested tables, key-value fields, and line items from invoices, bills of lading, and statements with coordinate-level evidence, multi-stage review, and direct API/webhook relay.
Coordinate-linked Bounding BoxesMulti-stage human reviewQuickBooks QBO & Native MCP
Engineered for document-heavy operations
Accounts payable & APFreight & logisticsPurchasing & ERPFinancial operationsHealthcare & legal
4-way ingestion channels
Capture documents anywhere. Process in one place.
Stop chasing email attachments, shared folders, and paper stacks. Bring all incoming files into a unified document inbox with automatic deduplication and provenance tracking.
Dedicated parser inboxes
Each parser gets a unique email address. Forward supplier emails to automatically capture PDFs and attachments.
Public guest upload portals
Send zero-login upload links (/upload/[token]) to vendors with custom required fields and regex checks.
Embeddable upload widget
Drop our responsive iframe widget (/embed/[token]) directly into your existing vendor portal in 60 seconds.
High-concurrency REST API
Stream multi-page PDFs, TIFFs, and Office files up to 50MB programmatically with webhook notifications.
Familiar from the first minute
Works like your document inbox, powered by precision intelligence.
Upload, email, or send files through an API. Each document keeps its immutable original scan, bounding-box evidence, and multi-stage approval history in one place.
- Unified multi-channel intake from email inboxes, partner upload links, or cloud sync
- Review only fields that need attention with side-by-side coordinate highlights
- Relay the exact approved result version downstream via Webhooks, CSV, or QuickBooks QBO
Document inbox
6 documents · 2 need review
Northwind_Invoice_1048.pdfNeeds review
Harbor_PO_2841.pdfNeeds review
Ocean_BOL_7783.pdfApproved
May_Statement.pdfApproved
Extracted data
2 checks flaggedVendorNorthwind Parts
Invoice #INV-1048
Invoice dateMay 24, 2026
Total$1,950.00
DescriptionQtyAmount
Professional services1$1200.00
Software license2$600.00
Support3$150.00
Beyond raw text extraction
An enterprise data pipeline built on verified truth.
Native layout understanding, coordinate-level DocumentIR, and sandboxed enrichment normalize into one reliable boundary before downstream business systems consume the result.
Provider-independent DocumentIR
Decouple from volatile model vendors. Document layout, coordinates, and tables normalize into one schema.
Ask AI schema builder
Drop a sample file to generate schema fields in seconds. Refine with 7 typed primitives and nested tables.
Deterministic data enrichment
Cross-reference external CSV catalogs to map vendor IDs, calculate math aggregations, and run secure JS scripts.
Direct relay & native MCP
Export OpenXML Excel, CSV, JSON, and QuickBooks QBO files instantly. Connect Claude Desktop and Cursor via /api/mcp.
From intake to delivery
A deterministic path for every mission-critical document.
No black boxes. Every file moves through explicit states, visible ownership, and an immutable audit trail.
1. Ingest & classify
Capture multi-page PDFs, TIFFs, or Office files up to 50MB via email, API, or widget. Prompt-guided classifiers identify types and route to target parsers.
2. Extract & validate
High-precision layout understanding extracts header fields and nested tables with calibrated confidence scores and validation rules on every value.
3. Review & relay
Reviewers verify flagged exceptions on side-by-side screens with bounding box evidence. Relay approved data via signed webhooks, Excel, CSV, or QuickBooks QBO.
Why ParseRelay
Built for the high-stakes decisions after extraction.
Reliable automation comes from validation, evidence, review policy, and controlled delivery — not from an unverified model response alone.
Evidence over guesswork
Every extracted value points directly to its bounding-box coordinates on the original document.
Versioned parser schemas
Fields, tables, and validation rules resolve to an explicit contract with atomic CAS revision control.
Multi-stage human review
Configurable approval gates with auditor assignments, line-item corrections, and full reason logs.
Automated decision rules
Set auto-approve and auto-reject rules based on mathematical reconciliation and confidence thresholds.
Enterprise BYOK & BYOS
Bring your own OpenAI/Claude/Gemini keys and private S3/R2 storage buckets for 100% data custody.
Agent-native with MCP
Native Model Context Protocol server (/api/mcp) enables Cursor, Claude Desktop, and Windsurf integration.
Industries we serve
Built for document-heavy operations
The same evidence, schema, review, and delivery model adapts to the operating documents each team depends on.
B2B distribution
Purchase orders, sales orders, commercial invoices, and proof of delivery.
Explore workflowAccounting & finance
Invoices, receipts, statements, and review-ready transaction data.
Explore workflowPurchasing & procurement
Supplier documents validated against expected references and totals.
Explore workflowFinancial operations
Statement tables and supporting documents normalized with evidence.
Explore workflowShared service teams
A consistent intake and review process across business units.
Explore workflowWho is ParseRelay for?
Give every owner the context they need
Operations, finance, engineering, and review teams work from the same versioned document record instead of disconnected handoffs.
AP manager
“I need invoice data I can verify before it reaches the ledger.”
Operations manager
“I need one queue for documents arriving from customers and suppliers.”
Logistics coordinator
“I need shipment references and charges captured without retyping PDFs.”
Bookkeeper
“I need consistent fields across many client and vendor formats.”
Data engineer
“I need stable schemas, idempotent jobs, and delivery evidence.”
Reviewer
“I need uncertainty, source context, and correction history in the same screen.”
Choose your use case
One product boundary, many document schemas
Every workflow still moves through intake, DocumentIR, extraction, validation, review policy, and controlled delivery.
Invoices & bills
- Vendor details
- Dates and references
- Totals and tax
- Line items
Bank statements
- Statement period
- Opening balance
- Transactions
- Closing balance
Shipping documents
- Consignee and shipper
- Container references
- Weights and packages
- Charges
Purchase orders
- Supplier
- PO number
- Requested items
- Delivery terms
Email attachments
- Message context
- Attachment routing
- Parser selection
- Thread evidence
Custom business forms
- Typed fields
- Repeatable tables
- Validation rules
- Delivery mapping
Calibrated review
Focus people on business uncertainty
Recognition probability is only one signal. Review policy can combine evidence, type and format validation, business rules, model agreement, document quality, and reviewed outcomes.
Review queue
Northwind_Invoice_1048.pdf
Total amountApproved
$1,950.00Totals reconcile
Purchase orderReview
PO-2841Reference not found
Vendor nameApproved
Northwind PartsEvidence aligned
Payment termsApproved
Net 30Expected format
Delivery paths
Relay data after approval
Exports and integrations should always identify the approved document and exact extraction-result version they deliver.
CSV & spreadsheets
Human-readable files for downstream review.
Signed webhooks
Retryable events protected against replay.
Business systems
Owned connector mappings tied to a schema version.
Safe retries
Idempotency prevents duplicate documents and side effects.
Approved result
Delivery policy
Downstream system
Operational trust
A visible record from source to delivery
Model runs, schema versions, transformations, reviews, approvals, and delivery attempts remain traceable instead of collapsing into one mutable output.
Original received
Immutable source and trusted tenant context.
Result versioned
Typed fields, tables, warnings, and evidence.
Decision recorded
Reviewer, corrections, policy, and timestamp.
Delivery attempted
Exact result version, retry state, and outcome.
Parser schemas & Ask AI
Make the expected result explicit.
Define 7 typed fields, repeatable tables, business rules, review thresholds, and delivery mappings in one versioned contract. Or let Ask AI infer the schema from a sample in seconds.
Explore an invoice workflowInvoice schema
Version 4 · Published (CAS)
invoice_numberTextRequired · Regex Matchinvoice_dateDateRequired · Format YYYY-MM-DDvendor_nameTextLookup ERP Supplier IDline_itemsTableRepeatable · 7 columnstotal_amountNumberMust reconcile sum(line_items)Predictable usage-based pricing
Start small, scale seamlessly with your volume.
All plans include 20 free processing credits on signup, full core feature access, and permanent credit refill packs that never expire.
Starter
For focused workflows and small operational teams.
$69/month ($49 billed annually)
Includes 20 free signup trial credits
- 500 Credits/mo ($0.138/page)
- 5 active parsers
- 3 team members
- PDF, TIFF, Office up to 50MB
- Dedicated email inboxes
- XLSX, CSV, JSON, XML & QBO export
Most Popular
Growth
For growing businesses managing mixed paperwork.
$199/month ($149 billed annually)
Includes 20 free signup trial credits
- 2,000 Credits/mo ($0.099/page)
- 25 active parsers
- 10 team members
- Document Classifiers included (0.1 credit/doc)
- Advanced nested table extraction
- CSV & cross-parser lookup tables
- Priority support queue
Business
For high-volume operations requiring zero-tolerance review.
$399/month ($299 billed annually)
Includes 20 free signup trial credits
- 5,000 Credits/mo ($0.079/page)
- 100 active parsers
- 25 team members
- Multi-stage human review workflows
- Configurable auto-approve & reject rules
- Isolated JS code sandbox post-processing
- 99.9% dedicated uptime SLA
Enterprise & Data Sovereignty
Need BYOK, BYOS, or custom volume?
Deploy with your own OpenAI/Claude keys, private S3/R2 storage buckets, SAML SSO/SCIM, and white-label agency branding.
Frequently asked
Questions before you relay
The product is engineered around a deterministic evidence pipeline. Here is how our architecture guarantees data accuracy and privacy.
How does ParseRelay eliminate hallucinations compared to generic LLMs?+
Generic LLMs generate text based on probability without spatial grounding. ParseRelay normalizes documents into a coordinate-level DocumentIR. Every extracted field is tied to a verified bounding box on the original document and validated against mathematical reconciliation rules.
How do Document Classifiers route mixed batches of paperwork?+
Our Document Classifiers use custom natural language instructions to inspect incoming files and identify whether a document is an invoice, purchase order, or bill of lading. It automatically routes the file to the corresponding parser at just 0.1 credits per classification.
What makes QuickBooks QBO export unique?+
Unlike generic CSV exports that require manual reformatting, ParseRelay can export validated data as native QuickBooks Web Connect (.QBO OFX 1.02 SGML) transaction files. You can import transactions directly into QuickBooks Desktop or Online with bank-matched fields.
Can we connect our AI assistants (Cursor, Claude Desktop, Windsurf) to ParseRelay?+
Yes. ParseRelay includes a native Model Context Protocol (MCP) server at /api/mcp. Developers and knowledge workers can use natural language prompts in their AI agents to list parsers, inspect document records, or approve review queues.
Can we run ParseRelay using our own cloud storage and AI keys?+
Yes. Enterprise plans support Bring Your Own Key (BYOK) for OpenAI, Anthropic, or Gemini, and Bring Your Own Storage (BYOS) for AWS S3, Cloudflare R2, or Google Cloud Storage. Your sensitive documents never touch third-party storage.
Stop rekeying documents. Start relaying verified data.
Create your first parser in under 2 minutes. Get 20 free processing credits with full feature access — no credit card required.