Airparser

@airparser list all my inboxes and show me which ones have documents waiting to be processed

Found 4 active inboxes in your Airparser account. **Vendor Invoices** (12 documents, last upload 2 hours ago) uses GPT-4 Vision for PDF extraction. **Customer Contracts** (3 documents) uses text extraction with a 15-field schema. **Expense Reports** (0 documents) is configured but empty. **Meeting Notes** (47 documents) has the highest volume this month. All inboxes are on your Pro plan with 8,200 parsing credits remaining this billing cycle.

This example demonstrates read-only discovery across the account. Requires a valid API key with inbox-read scope. The MCP does not auto-parse documents—it surfaces what Airparser has already processed. Useful for auditing which document families you're tracking and where parsing bottlenecks exist.

@airparser create a new inbox called 'Supplier Quotes' that uses vision extraction to pull item descriptions, quantities, and unit prices from PDF quotes

Created inbox **Supplier Quotes** (ID: `ibx_9k3mq7`) with GPT-4 Vision engine. The extraction schema includes three fields: `item_description` (text), `quantity` (number), and `unit_price` (currency). You can now forward supplier quote PDFs to the inbox's unique email address or upload via API. Each document will consume ~50 parsing credits depending on page count and complexity. Schema can be refined after reviewing initial extractions.

This write operation provisions a new parser and consumes account resources. Vision-based extraction works well for scanned or image-heavy PDFs but costs more credits than text extraction. You'll need to iterate on the schema—first-pass field definitions rarely capture every edge case in real supplier documents.

@airparser get the last 5 documents from my 'Vendor Invoices' inbox and summarize any invoices over $5,000 that are due within 10 days

Retrieved 5 most recent documents from **Vendor Invoices**. Two invoices exceed $5,000 and are due soon: **Acme Supply Co.** ($7,340 due March 18) for industrial components, and **Metro Logistics** ($12,800 due March 15) covering February shipping. Both documents parsed successfully with high confidence scores. The other three invoices are under $3,000 or due after March 25. All extraction timestamps are within the last 48 hours, so data is current.

This synthesis example combines the MCP's document-retrieval tool with the AI's filtering and summarization logic. The MCP returns structured extraction output (amounts, dates, vendor names); the AI applies your business rules. Accuracy depends on Airparser's schema quality—if the inbox wasn't trained to extract due dates, this query will fail silently.

When Airparser beats manual entry for recurring vendor invoices

A three-person accounting firm handles 40-60 vendor invoices a week across five clients. Each invoice arrives as a PDF attachment with different layouts—utilities, SaaS subscriptions, contractors. Airparser wins here because you create one inbox per vendor type, train the extraction schema once using the vision engine, then forward invoices to the inbox email. The Get Document and List Documents tools let your team pull parsed line items into QuickBooks or Xero without re-keying. The break-even point is around 15 invoices a month per vendor; below that, manual entry is faster than schema tuning. If your invoices are truly one-off (new vendors every week), the setup tax outweighs the benefit. For stable vendor rosters, this MCP turns email into structured data with two minutes of schema work up front.

Airparser for high-volume attachment parsing in support queues

A 12-person SaaS support team gets 200 tickets a day, and 30 percent include screenshots, logs, or error PDFs. The team needs to route tickets by severity and product area, but agents waste 90 seconds per attachment deciding if it's urgent. Airparser's text and vision engines can parse common attachment types—stack traces, browser console logs, annotated screenshots—into structured fields that feed your triage rules. Create one inbox for logs, one for screenshots, use List Documents to poll new arrivals every five minutes, then tag tickets in Zendesk or Intercom based on parsed severity markers. The constraint: if your attachment formats change weekly (custom client exports, ad-hoc spreadsheets), you'll spend more time updating schemas than you save. This MCP shines when 70 percent of attachments fit three or four repeating templates.

When Airparser accelerates NDA and MSA review cycles

A two-lawyer startup reviews 10-15 NDAs and MSAs a month before signing with vendors or partners. Each contract is a 6-20 page PDF with standard clauses buried in different sections—liability caps, termination terms, indemnification language. Airparser's vision engine can extract those clauses into a comparison table if you build one inbox per contract type and define the schema around your must-check fields. Use Get Inbox to confirm the schema matches your review checklist, then forward signed PDFs to the inbox and pull parsed clauses with Get Document. The trade-off: if every contract is bespoke (custom riders, foreign jurisdictions, heavily redlined), the extraction accuracy drops below 80 percent and you're back to manual review. For repeating contract forms from the same three law firms, this MCP cuts first-pass review from 20 minutes to two.