PDF Vector vs Affinda: 2026 Feature Comparison

PDF Vector vs Affinda comes down to a platform versus an API. Affinda Extract is a document automation platform for operations teams: it ingests, splits, classifies, extracts, validates against business rules, routes exceptions to human reviewers and writes the result into ERPs and other core systems. PDF Vector is a developer API: you send a document with a JSON Schema and get JSON back, or get Markdown, or an answer to a question, at a published price per page. If you need review queues, evidence on every field and self-hosting, Affinda is the stronger fit. If you are building document features into your own software or automations and want self-serve pricing, PDF Vector is. Affinda facts below come from its own pages, checked November 2026.

Quick comparison

	PDF Vector	Affinda
Product shape	REST API, TypeScript SDK, MCP server, Zapier/Make/n8n actions	Affinda Extract platform (web app, Agent, API, client libraries, webhooks); separate Resume Parser API
How extraction is defined	Prompt + JSON Schema in each request	Document types and fields configured in the platform, in conversation with the Affinda Extract Agent
Document workflow steps	Parse, Extract, Ask per document	Ingestion, splitting, classification, extraction, validation, integration
Human review	Not included; build it in your app or automation	Review UI, embeddable in your application; exceptions routed on failed validation rules
Source evidence	Not returned (no coordinates)	Evidence linked to every extracted value
Data matching	In your own code	Matching against ERP, CRM or master data
Specialized documents	Invoice, identity and bank statement endpoints	Configurable per document type; dedicated Resume Parser with occupation and skill taxonomies
Research content	Academic search across eight sources, paper parsing, citations, grants	Not offered
Integrations	Zapier, Make, n8n, MCP clients	Named connectors incl. NetSuite, SAP Business One, Xero, Salesforce, Power Automate, n8n, Pipedream; an AI connector for agents
Deployment	Shared instance (US); Enterprise dedicated instances in UK, EU, US, Japan, Australia or Asia Pacific	Multi-tenant SaaS, single-tenant, customer VPC or self-hosted
Compliance stated	Files are not used for training	ISO 27001:2022, SOC 2 Type 2, GDPR; HIPAA-compliant deployment available
Pricing	Free 100 credits/month; Basic $25/month (3,000 credits); Pro $97/month (100,000 credits); Enterprise $299/month	Extract: per page, quoted (monthly in arrears or annual commitment), 2-week trial with 200 credits; Resume Parser: from US$0.10/document

Sources: Affinda Extract, pricing, integrations, Resume Parser.

What Affinda does well

Affinda's own description of Affinda Extract (formerly Affinda Platform) is "one platform, six stages": ingestion from inboxes, portals and storage; splitting bundled packs; classification; extraction; validation; and integration. Its use cases are operational: lending teams clearing 60-to-120-page application bundles, insurance teams validating ACORD packs, AP teams matching invoices to purchase orders, onboarding teams checking IDs and proof of address.

The strengths that matter in those settings:

Evidence and auditability. Every extracted value links back to the source, review decisions have audit trails, and corrections are reused by the workflow as instructions rather than by retraining the model.
Validation before export. Natural-language validation rules and master-data matching check each record; documents that pass can go straight through, and the rest wait in a review queue. The review UI can be embedded in your own product.
Configuration in conversation. You describe documents, rules and destinations to the Affinda Extract Agent and it proposes fields, validation logic, reviewer routes and integration endpoints for you to approve.
Deployment and compliance. SaaS, single-tenant, customer VPC or self-hosted; ISO 27001:2022 and SOC 2 Type 2; a HIPAA-compliant deployment option; data residency and SSO as paid add-ons.
Resume parsing. The Affinda Resume Parser accepts PDF, Word, text, HTML, images, ODT and Excel files in 50+ languages, maps job titles to occupation codes and normalizes skills to taxonomies, hosted or self-hosted, with published pricing.

If those are requirements, Affinda is solving more of your problem than an API can.

What PDF Vector does well

PDF Vector keeps the integration surface small: four document families (general documents, invoices, identity documents, bank statements), each with the same three operations.

Document Extract, Invoice Extract and Identity Extract take a prompt and a JSON Schema and return data in exactly that shape. Change the schema and the output changes on the next request: no configuration step, no training.
Document Parse returns the whole document as Markdown (and HTML on max), optionally per page, for LLM, search and RAG pipelines.
Ask answers a question about a document, such as "Does this contract renew automatically?"
Academic APIs: Academic Search across Semantic Scholar, PubMed, arXiv, Google Scholar, ERIC, Europe PMC, OpenAlex and Crossref, plus paper fetch and parse, citation suggestions, citation graphs, similar papers and grant search.

Formats: PDF, Word, Excel, PowerPoint, CSV, HTML, RTF, OpenDocument, EPUB and text on every tier; images (PNG, JPG, TIFF, BMP, HEIF) on pro, max or auto. Files up to 40 MB and 30 pages on nano, mini and pro, and up to 500 MB and 2,000 pages on max and auto.

The same invoice workflow in each

In Affinda Extract, you would describe your invoice documents, the fields and validation rules (for example, "the total must match the open PO within 1%") and the ERP destination to the Agent, approve its configuration, and let the platform run: documents that pass go to the ERP, the rest appear in the review queue with the source evidence highlighted.

With PDF Vector, you call the API and put the rules in your code. A TypeScript example that extracts an invoice and checks it before export:

import { createClient } from "@pdfvector/client";

const client = createClient({ apiKey: process.env.PDFVECTOR_API_KEY });

type Invoice = {
  vendor: string;
  invoiceNumber: string;
  poNumber?: string;
  total: number;
  lineItems: { description: string; amount: number }[];
};

const result = await client.invoice.extract(
  {
    url: "https://example.com/invoices/inv-0892.pdf",
    prompt: "Extract vendor, invoice number, purchase order number, total and every line item amount",
    schema: {
      type: "object",
      properties: {
        vendor: { type: "string" },
        invoiceNumber: { type: "string" },
        poNumber: { type: "string" },
        total: { type: "number" },
        lineItems: {
          type: "array",
          items: {
            type: "object",
            properties: {
              description: { type: "string" },
              amount: { type: "number" },
            },
            required: ["description", "amount"],
          },
        },
      },
      required: ["vendor", "invoiceNumber", "total", "lineItems"],
    },
    model: "auto",
  },
  { context: { documentId: "inv-0892" } },
);

const invoice = result.data as Invoice;
const lineSum = invoice.lineItems.reduce((sum, item) => sum + item.amount, 0);
const problems: string[] = [];
if (Math.abs(lineSum - invoice.total) > 0.01) problems.push(`line items sum to ${lineSum.toFixed(2)}, total is ${invoice.total}`);
if (!invoice.poNumber) problems.push("missing PO number");

if (problems.length === 0) {
  // write to your ERP or accounting system
} else {
  // send to a person: a Slack message, a ticket or your own review screen
}
console.log(`${result.pageCount} pages, ${result.credits} credits`, problems);

This example compares line items with the total and leaves out tax and shipping lines; adapt the check to your invoices. More checks are covered in How to Validate Extracted Invoice Data. The trade-off is clear: with PDF Vector you write and own the validation and review step; with Affinda you configure them.

Pricing compared

PDF Vector publishes every price. Credits are charged per page and depend on the operation and tier:

Operation	nano	mini	pro	max
Document Parse	1	2	4	8
Document Ask / Extract	2	4	8	16
Invoice / Identity / Bank Statement Parse	–	–	6	10
Invoice / Identity / Bank Statement Ask / Extract	6	10	14	18

Worked examples:

A 3-page invoice on Invoice Extract mini: 3 × 10 = 30 credits. The Basic plan's 3,000 credits cover 100 such invoices (3,000 ÷ 30).
2,000 two-page invoices a month on nano: 2,000 × 2 × 6 = 24,000 credits, within the Pro plan's 100,000 ($97/month).
A 40-page loan agreement parsed to Markdown on max (the 30-page cap applies to nano, mini and pro): 40 × 8 = 320 credits.

Plans: Free (100 credits/month, no card), Basic $25/month ($23 billed annually) for 3,000 credits, Pro $97/month ($89 billed annually) for 100,000 credits, Enterprise $299/month with a dedicated managed instance. Credits reset monthly and do not roll over.

Affinda prices Affinda Extract by pages processed, with no seats or feature tiers, and includes all core capabilities (custom document types, splitting, classification, validation, review UI, API, webhooks) in standard usage. The pricing page offers a two-week trial with 200 credits, then monthly billing in arrears or an annual commitment where overage is charged at your committed rate; per-page rates are quoted, not published. On-premises deployment, data residency controls, SSO, HIPAA support, dedicated throughput and premium support are paid add-ons. The Resume Parser has published prices: US$0.10 per document self-serve, annual hosted plans from US$3,600 for 66,000 documents, and self-hosted plans from US$12,000 a year for 500,000 documents (resume parser pricing).

So you can estimate PDF Vector costs from a spreadsheet; for Affinda Extract you need a quote. That is not a sign either is cheaper at your volume: compare a real quote with the arithmetic above.

Data, security and deployment

Affinda lists ISO 27001:2022 certification, SOC 2 Type 2 and GDPR compliance, a HIPAA-compliant deployment, encryption at rest and in transit, MFA for every user, AWS data centres in multiple regions, and deployment as multi-tenant SaaS, single-tenant, customer VPC or self-hosted. It states customer data is not used to retrain underlying models.
PDF Vector runs a shared instance in the US. Enterprise customers get dedicated managed instances that can run in the UK, EU, US, Japan, Australia or Asia Pacific. Files are not used for training. Self-hosting is not offered.

If your security review requires specific certifications or on-premises deployment, Affinda meets more of those requirements today.

Choose Affinda if

Exceptions need a person and an audit trail. Lending, insurance claims, AP with PO matching, KYC onboarding: anywhere a wrong value has a cost and reviewers need the source evidence in front of them.
Documents arrive as bundles. Splitting and classifying 100-page packs before extraction is built in.
The data must land in an ERP, LOS or policy system and you want the platform to own that integration.
You need self-hosting, a VPC deployment or HIPAA.
You are parsing resumes for an ATS and want taxonomy-normalized skills and occupations.

Choose PDF Vector if

You are building a product or internal tool and want extraction defined in code, with the schema versioned next to the rest of your application.
You want to start today at a known price. Sign up, get an API key, and estimate cost per page before you commit.
Your documents vary widely: contracts, reports, statements, forms and spreadsheets through one API, with Markdown for LLM pipelines and Ask for quick answers.
Your team works in Zapier, Make or n8n, or in an MCP client such as Claude or Cursor.
Research content matters. Academic search, paper parsing and citation tools sit in the same account. See the literature review use case.

How to evaluate both

Take 30 to 50 real documents per type, including scans and your hardest layouts.
Define the target fields once as a JSON Schema. Use it directly in PDF Vector and as the field list when configuring Affinda.
Score accuracy per field, and time how long a reviewer spends on each exception.
Price your monthly pages with PDF Vector's table above and ask Affinda for a quote at the same volume.
Run one document end to end into the system that will receive the data.

If neither fits, Best Affinda Alternatives in 2026 covers Mindee, Veryfi, Nanonets, Azure Document Intelligence and Eden AI.