Parseur

Document Processing

Turn unstructured documents into structured agent inputs. Parseur extracts data from invoices, emails, and PDFs. Two trigger modes let agents act instantly when a document is processed or poll on a schedule.

2 triggers 6 actions available

On Document Processed (Realtime) fires when vendor invoice arrives in Parseur mailbox

Structured data arrives with vendor name, invoice number, amount, line items, due date

Agent cross-references vendor against approved vendor list in Acumatica

Agent checks for duplicate invoices by reference number

Agent creates bill in Acumatica and attaches original PDF document

AP team notified via Slack for bill approval

Invoices with amounts 3x higher than vendor average paused for reviewer verification

What This Integration Enables

Parseur sits at the front of document-driven workflows. It processes incoming documents and outputs structured JSON: invoice numbers, amounts, dates, line items, contact information. FlowRunner agents receive those structured payloads via trigger and feed them into every downstream system: ERP, CRM, accounting, storage. The human who used to enter that data gets time back.

Without FlowRunner

Manual invoice entry Each vendor invoice takes 8 minutes of manual data entry

Inconsistent format handling Invoices from different vendors require different entry approaches

Silent extraction failures Failed extractions sit unnoticed until someone checks the queue

With FlowRunner

15-second automated processing Three invoices that took 45 minutes of manual entry now process in under a minute total

Format-agnostic extraction Parseur normalizes any invoice format into consistent JSON for downstream systems

Immediate failure escalation Extraction failures routed to ops team via Slack the moment they occur

Use Case Scenarios

Automated Invoice Processing

Vendor invoices arrive by email to a Parseur mailbox. On Document Processed (Realtime) fires. The extracted data arrives: vendor name, invoice number, amount, due date, line items. The agent cross-references the vendor against the approved vendor list in Acumatica. It checks for duplicate invoices by reference number. If it passes both checks, it creates a bill in Acumatica, attaches the original document, and routes for approval via Slack. Three invoices that each took 8 minutes to enter manually now process in 15 seconds each.

Purchase Order Confirmation Extraction

Customers email purchase orders in varying formats: some are Word documents, some are PDFs, some are structured emails. Parseur normalizes them into a consistent JSON structure. On Document Processed fires. The agent creates the sales order in NetSuite from the extracted line items, sends an order confirmation email, and logs the PO in the CRM. No one in ops touches a keyboard.

Failed Extraction Review

The agent receives an On Document Processed trigger with a failed status flag. The document extraction failed: it could be a non-standard format or a scanned image with poor quality. The agent routes the failure to the ops team via Slack: "Document extraction failed for [filename]. View the document and reprocess, or enter manually?" The team handles the exception. The document does not disappear into a queue unnoticed.

Human-in-Loop Highlight

Not every extraction is perfect. When Parseur processes a document and the extracted amount does not match expected ranges, the agent pauses before pushing the data downstream. It sends the extracted data to a reviewer via Slack: "Extraction from [vendor] invoice: $94,200. This is 3x higher than this vendor's typical invoice. Verify before I create the bill?" The reviewer confirms or corrects the amount. The agent proceeds with verified data, not an unchecked extraction.

Agent processes routinely

Detects exception requiring judgment

Clear match Continues automatically

Ambiguous Routes to human via Slack

Human decides

Agent resumes with decision

Agent Capabilities

6 actions

Mailbox Management

Create Mailbox Creates a new Parseur mailbox for document collection. Used in workflow setup: when a new document type or source needs its own processing channel, the agent creates the mailbox automatically.

Document Operations

Upload Document Submits a document to a Parseur mailbox for parsing. Used when documents arrive through channels other than email: an agent retrieves a PDF from S3 or a shared folder and pushes it into Parseur for extraction.
Get Parsed Data Retrieves the structured output from a processed document. Used when agents need to fetch extraction results without waiting for a trigger, typically in polling or batch workflows.
List Documents Returns the documents in a mailbox with filtering options. Used in monitoring, reporting, and reprocessing workflows.
Reprocess Document Re-runs parsing on a document using a different or updated template. Used when an initial extraction was incorrect and the document needs to be re-extracted after template correction.
Delete Document Removes a document from the mailbox. Used in cleanup and data governance workflows.

Triggers

2 triggers

Event Triggers

On Document Processed (Realtime) Fires via webhook the instant Parseur finishes processing a document. Zero delay between document processing and agent action. The extracted data payload arrives with the trigger: document ID, processing status, the structured parsed data (invoice numbers, amounts, dates, vendor names), mailbox ID, filename, and timestamp. This is the primary trigger for high-volume AP and data intake workflows.
On Document Processed (Polling) Checks for newly processed documents at a configurable interval (minimum 30 seconds). Used when webhook connectivity is not available or when batch processing cadence is preferred. Includes the option to capture failed processing attempts. Both triggers support mailbox selection and an "Include Failed" option to surface extraction failures for review.

Start building with Parseur

$100 in credits. No card required. Connect in minutes.

Start Building Free Book a demo