> This page is for version v2026-02-09 (default).
> For other versions, use one of these documentation indexes:
> - v2026-02-09 (default): https://docs.extend.ai/2026-02-09/llms.txt
> - v2025-04-21: https://docs.extend.ai/2025-04-21/llms.txt
> - v2024-12-23: https://docs.extend.ai/2024-12-23/llms.txt

> ## Documentation Index
> Fetch the complete documentation index at: https://docs.extend.ai/llms.txt
> Use this file to discover all available pages before exploring further.
>
> ## API version
> The current API version is `2026-02-09`, served at the site root (no version prefix in URLs).
> If this page URL contains `/2025-04-21/` or `/2024-12-23/`, you are reading an older API version.
> Prefer the current docs at https://docs.extend.ai/llms.txt unless the user explicitly needs that older version.
> Do not treat older-version pages as the source of truth for new integrations.

> How Extend's credit-based billing works: what parse, extract, classify, split, and edit runs cost, how Performance vs Light modes change price, and plan rates.

# How Credits Work

With Extend, your cost depends on **what you run** (e.g. Parse, Classify, Split, Extract, Edit Runs, Form Detection), **how you run them** (e.g. Performance vs. Light modes), and your **plan's credit rate** (Pay as you go, Scale, or Enterprise).

Instead of fixed fees, you only pay for the specific processors you need, at the level of performance you require.

* Credits are Extend's universal billing unit.
* Each action (parse, classify, extract, split, edit, form detection) consumes a base amount of credits per page, with optional surcharges for advanced features.
* Your plan determines the cost per credit, and plans with higher volumes get **cheaper per-page pricing**.
* You stay in control with alerts, top-ups, and real-time dashboards.

> **Note**
>
> **Plans at a glance.** Pay as you go is free self-serve: 10,000 free credits, then \$0.0125 per credit, with no contract, monthly minimum, or sales call. Scale (\$500/month, 50,000 credits included, \$0.01 per additional credit) is an optional volume tier. Enterprise pricing is custom. See [pricing](https://www.extend.ai/pricing).
>
> **Light vs Performance.** Light modes cost 4× less than Performance modes for Parse and Split (0.5 vs 2 credits/page) and about 4.3× less for Extract (0.7 vs 3 credits/page). Choose the mode per processor.

---

## 🧾 Base Pricing (Performance)

| Action         | Credits/page | Pay as you go | Scale  | Enterprise |
| -------------- | ------------ | ------------- | ------ | ---------- |
| **Parse\*\***  | 2            | \$0.025       | \$0.02 | Custom     |
| Extract        | 3            | \$0.0375      | \$0.03 | Custom     |
| Split          | 2            | \$0.025       | \$0.02 | Custom     |
| Classify       | 1            | \$0.0125      | \$0.01 | Custom     |
| Edit Filling   | 1            | \$0.0125      | \$0.01 | Custom     |
| Form Detection | 2            | \$0.025       | \$0.02 | Custom     |

## 🧾 Base Pricing (Light)

| Action        | Credits/page | Pay as you go | Scale   | Enterprise |
| ------------- | ------------ | ------------- | ------- | ---------- |
| **Parse\*\*** | 0.5          | \$0.00625     | \$0.005 | Custom     |
| Extract       | 0.7          | \$0.00875     | \$0.007 | Custom     |
| Split         | 0.5          | \$0.00625     | \$0.005 | Custom     |
| Classify      | 0.3          | \$0.00375     | \$0.003 | Custom     |

## 🧾 Base Pricing (Auto Parsing)

The **Auto** parse engine (`parse_auto`) routes work per page: simple pages are parsed with the lightweight pipeline while complex pages get performance-level treatment. Each page is billed by the complexity of work it required:

| Page tier   | When it applies                                  | Credits/page | Pay as you go | Scale   | Enterprise |
| ----------- | ------------------------------------------------ | ------------ | ------------- | ------- | ---------- |
| Light       | Simple pages handled by the lightweight pipeline | 0.5          | \$0.00625     | \$0.005 | Custom     |
| Performance | Pages routed to Performance models               | 2            | \$0.025       | \$0.02  | Custom     |

The per-tier charges are itemized on the run response in `usage.breakdown`, including which pages fell into each tier:

```json
{
  "charges": [
    { "product": "parse_auto_light", "unit": "page", "quantity": 3, "credits": 1.5, "pages": [1, 2, 3] },
    { "product": "parse_auto_performance", "unit": "page", "quantity": 1, "credits": 2, "pages": [4] }
  ]
}
```

## 🧾 Base Pricing (Operator)

| Action  | Credits/page | Pay as you go | Scale  | Enterprise |
| ------- | ------------ | ------------- | ------ | ---------- |
| Extract | 5            | \$0.0625      | \$0.05 | Custom     |

Most Operator-1 runs stay at **5 credits/page** for extraction, including large table extractions and other long documents. Rare overages can apply when a short document requires unusually intensive work, such as interpreting a complex diagram or extracting information from a few extremely dense pages.

Parsing is billed separately. Operator runs are exempt from Review Agent, array-strategy, and large-schema surcharges.

🤖 \*\* Parse runs automatically when Extract, Split, or Classify processors are used.

💡 Edit pricing:

* Edit Filling (`/edit`) costs **1 credit per page**. If no schema is passed to Edit Filling, it runs Form Detection and both charges will apply (so 3 total credits per page). If you provide a schema, only the Edit Filling charges apply (1 credit per page).
* Form Detection (`/detect_form` or `/form_detection_runs`) costs **2 credits per page**. The deprecated `/edit_schemas/generate` endpoint uses the same rate.

---

## 🔧 Optional Surcharges & Multipliers

The following optional features add surcharges on top of the base credit rates when enabled.

### Review Agent

The Review Agent automatically flags results for quality assurance and adds a more reliable confidence score. It is available for both extraction and classification processors.

| Feature                     | Surcharge         |
| --------------------------- | ----------------- |
| Extraction Review Agent     | +1 credit/page    |
| Classification Review Agent | +0.5 credits/page |

### Agentic Corrections (Parsing)

When enabled, agentic correction features use a vision language model (VLM) to review and correct OCR results during parsing.

| Feature                  | Surcharge      |
| ------------------------ | -------------- |
| Agentic Text Correction  | +1 credit/page |
| Agentic Table Correction | +1 credit/page |

**Note:** When enabled for a document, agentic corrections only applies/charges for pages where the agentic correction was triggered (e.g. low confidence text blocks, tables, etc.).

### Priority Parsing

**Priority parsing** guarantees faster p90+ parse results on the light parsing pipeline. Reach out to the Extend team if you need to enable this for your account.

| Feature          | Multiplier | Applies to      |
| ---------------- | ---------- | --------------- |
| Priority Parsing | 2x         | Parsing credits |

### Array Strategy Multiplier (Extraction)

**Large Array Max Context** does multiple passes through the entire document to maximize accuracy for complex array extraction, doubling the base extraction credit cost.

| Feature                 | Multiplier | Applies to         |
| ----------------------- | ---------- | ------------------ |
| Large Array Max Context | 2x         | Extraction credits |

### Advanced Excel Parsing

Advanced Excel parsing uses LLM-based layout detection for better table separation and header recognition in complex spreadsheets. Billed per cell rather than per page. Does not include empty cells.

| Feature                | Rate                 |
| ---------------------- | -------------------- |
| Advanced Excel Parsing | 3 credits/1000 cells |

Excel files parsed in `basic` mode, or in `advanced` mode with tables output as `csv`, are not charged.

---

## 🆓 When Parsing Is Not Charged

Parse credits apply to files that go through the page-based pipeline: PDFs, images, and document formats that Extend converts to PDF before parsing (such as DOC, DOCX, PPT, and PPTX). Some file types are parsed directly from their text content instead, and those runs are recorded at **0 credits**:

* CSV, plain text, RTF, Markdown, and XML
* HTML and email (EML) files when they are parsed as-is rather than converted to PDF
* Excel files in `basic` parsing mode (see [Advanced Excel Parsing](#advanced-excel-parsing) for the per-cell rate)

---

### Example Scenario:

💡 **Formula:** `base credits per page × your plan's per-credit rate = $/page` (plus any applicable surcharges)

* Processing a 10-page document through a Splitting & Extraction workflow on Performance mode, would consume **70 credits** in total, itemized as follows:
  * 20 credits for parsing (10 pages \* 2 credits)
  * 20 credits for splitting (10 pages \* 2 credits)
  * 30 credits for extraction (10 pages \* 3 credits)

* If the same extraction run also has the **Review Agent** enabled, it would add 10 credits (10 pages \* 1 credit), for a total of **80 credits**.

* If the parsing step used the **race pipeline strategy** and had **agentic text correction** trigger on 5 pages and **agentic table correction** trigger on 3 pages, parsing would cost 48 credits (10 \* 2 \* 2 for race + 5 \* 1 for text correction + 3 \* 1 for table correction) instead of 20.

### Total cost by operation and mode

Because Parse runs automatically before Extract, Split, and Classify, the all-in cost per page is the sum of the operation and its parse. The table below gives the total at Pay as you go rates (\$0.0125/credit) with no surcharges enabled.

| Workload (per page)          | Light credits (total) | Light \$/page | Performance credits (total) | Performance \$/page |
| ---------------------------- | --------------------- | ------------- | --------------------------- | ------------------- |
| Parse only                   | 0.5                   | \$0.00625     | 2                           | \$0.025             |
| Extract (+ automatic Parse)  | 0.7 + 0.5 = 1.2       | \$0.015       | 3 + 2 = 5                   | \$0.0625            |
| Split (+ automatic Parse)    | 0.5 + 0.5 = 1.0       | \$0.0125      | 2 + 2 = 4                   | \$0.05              |
| Classify (+ automatic Parse) | 0.3 + 0.5 = 0.8       | \$0.01        | 1 + 2 = 3                   | \$0.0375            |

Worked monthly examples at Pay as you go rates:

* **10,000 pages/month, Light Extract:** 12,000 credits ≈ **\$150**. The first 10,000 credits are free on a new account.
* **10,000 pages/month, Performance Extract:** 50,000 credits ≈ **\$625**. On Scale (\$500/month with 50,000 credits included) the same workload is covered by the plan's included credits.
* **100,000 pages/month, Light Parse for RAG:** 50,000 credits ≈ **\$625** on Pay as you go, or **\$500** on Scale.

Mix modes per processor: for example, Performance Parse for scanned pages that feed extraction and Light Parse for born-digital pages that only feed retrieval.

---

## Previewing and inspecting costs

Extend gives you visibility into credit usage at every stage — while you're configuring a processor, after each run completes, and across your account over time.

### In the config builder (before you run)

When configuring a processor, the builder displays the per-page credit cost for your current settings — including any surcharges from features like the Review Agent or agentic corrections — so you can see exactly how each option affects pricing before you save.

![Config Builder Cost Preview](/_fern-img/537393f774d4ae302150dc3d4865ce02e2bc36f36e7ce802e293d3a0dd54afa6.webp)

### On a run response (after you run)

Every run response includes a `usage` object so you can see exactly what drove its cost. `usage.credits` is the billed amount, `usage.totalCredits` rolls up any work this run triggered (e.g. parsing kicked off by an extract), and `usage.breakdown` lists each contributing run. Each breakdown entry can include a `charges` array that itemizes the credit usage — base processor tier, surcharges like the review agent or agentic corrections, and (for page-scoped add-ons) the exact page numbers that incurred the charge.

For the full shape, see [`usage` on the extract run response](/api-reference/endpoints/extract/get-extract-run#response.body.usage).

### In the dashboard (over time)

See credit balance and usage trends in real time.

![Dashboard Usage](/_fern-img/353e236a818cc20379306b76aca6e18352ac03059a8b81eeb8cac6751ad05d63.webp)

Set low balance notifications and top-up preferences to stay ahead of usage.

![Alerts Settings](/_fern-img/74ac2c21e01afb63f00b7842e46e7cbc353c2f7b32f0e79c8c5f334812e33e7a.webp)

### Annual plan usage

If you're on an annual plan, the **Billing** page shows how your usage is tracking against your plan — including usage to date, how much remains, the days left in your term, and a chart of your usage over time. Depending on your plan, usage is shown either in dollars against your annual amount, or in credits remaining against a prepaid balance.

You can also set alerts to be notified when your usage reaches a chosen percentage of your plan.

![Annual Plan Usage](/_fern-img/21d5609d763fffa70bb4c784befd8e85eb485c7b6033af7eb6e83894062c08c6.webp)

---

## Customize Your Attributions

You can categorize your workflow, processor, and parser runs for better billing and usage tracking by including usage tags in your run metadata. This allows you to filter and visualize credit usage in the dashboard based on different dimensions like environment, team, or customer.

### Adding Usage Tags

Include usage tags in your run metadata using the `extend:usage_tags` key with an array of string values:

```json
{
  "extend:usage_tags": ["production", "use_case_1"]
}
```

### Tag Requirements

**Allowed characters:** Only alphanumeric characters (a-z, A-Z, 0-9), hyphens (-), and underscores (\_)

**Removed characters:** All special characters and symbols are stripped out, including:

* Spaces
* Punctuation (!, @, #, \$, %, ^, &, \*, etc.)
* Quotes and brackets
* Any Unicode/emoji characters

### Filtering in the Dashboard

The credits dashboard provides multiple filters to help you analyze your usage:

* **Usage Tags**
* **Processors**
* **Workflows**
* **API Keys**
* **Charge Types** (Parsing, Classification, Extraction, Splitting, Editing, Agent usage)

You can combine multiple filters to create detailed views of your usage patterns.

You can also group the dashboard by **Charge Type**, **Processor**, **Workflow**, or **API Key** to see a breakdown of credit usage along that dimension.

---

# Optimizing your Credit Usage

Looking for ways to improve workflow efficiency, or curious if Performance or Light mode is right for you? Book a consultation with the team [here](https://meetings.hubspot.com/mike-dombrowski/extend-intro-call)!