Content Operations: how units are measured

A Content Operation is a unit of work the Content Operations Agent performs on Contentful entities and actions. Contentful entities include entries, assets, and content types. Operations are measured as write operations or research operations. Each qualifying entity or action counts once,not once per field.

A write operation counts when the Agent successfully creates, updates, deletes, or archives a Contentful entity, changes a release, or invokes an App Action. The table shows which actions count.

Each qualifying entity or action counts once, regardless of how many fields are changed. AI Action word-based usage is explained separately below.

The Agent may retrieve details about Contentful entities to understand their content or structure before answering a question, finding information, or carrying out a requested task. A research operation counts only when the full details of a particular Contentful entity are loaded into context by the Agent. Searches and list actions—such as listing tags or content types—do not count as research operations. Each entity whose details are retrieved counts as one research operation. Retrieving information does not itself change the entity; if the Agent later updates or archives it, that action is counted separately as a write operation.

Examples of research operations:

• One Contentful entry’s details retrieved. 

• One Contentful asset’s details retrieved. 

• One Contentful Content type schema retrieved. 

Operations are counted per qualifying entity or action, not per field changed.

For example, if the agent changes three fields within one entry—including its status—that counts as one write operation. If it makes the same changes to three entries, that is three write operations. If the agent retrieves one entry’s details and then updates that entry, that is two operations: one research operation and one write operation.

Examples of Content Operations

These examples show how reads and writes add up. Searches and list actions do not count as research operations.

Bulk update

Prompt: “On this sales-pack page, go through every image component and update three fields: title, description, and category.”

The Agent performs two research operations: one to load the page and its list of components, and one to load a component and view its fields. It then makes 90 qualifying updates. That is two research operations plus 90 write operations, for 92 Content Operations. Changing three fields in one update does not make it three operations.

Build a page

Prompt: “Create a page with a text module, a text-and-image module, a headline module, an image module, a video module and a story slider with 5 slides.”

The Agent retrieves 8 content type schemas and creates 12 entries plus 1 release. That is 8 research operations plus 13 write operations, for 21 Content Operations.

Clean up stale content

Prompt: “Show me content that hasn't been updated in 6 months. Unpublish all of it, and archive the entries for the product launch sales campaign.”

The Agent opens 5 entries to confirm them, then unpublishes 91 entries and archives 5 entries. That is 5 research operations plus 96 write operations, for 101 Content Operations.

Content audit

Prompt: “Go through every product page published in the last two weeks and flag any Danish title that's wrong: English text, typos, or product details that don't match.”

The Agent opens 20 entries and retrieves 3 content type schemas, then returns a report without changing anything. That is 23 research operations and no write operations, for 23 Content Operations.

AI Action usage is measured by the words the AI Action produces in a Contentful entity. For example, when the Agent translates an entry using an AI Action, the generated words are measured as AI Action usage.

This word-based usage is separate from Content Operations. Running an AI Action on an entry will count the word output from the Action and those words will consume from the same pool of Consumption Units as the Content Operations Agent. You will not be double charged since the agent running the AI Action does not count as a Content Operation.  If the Agent performs other qualifying actions on the entity—for example, changing another field or adding tags—those actions are counted under the write-operation rules above. A workflow can therefore include both AI Action usage, measured by generated words, and Content Operations.

Consumption Units (CUs) measure usage of Contentful AI capabilities. AI Actions use CUs based on the words they generate. Content Operations also use CUs: each Content Operation uses 0.25 CUs, so four Content Operations use one CU.

Your Contentful plan includes a CU allowance for AI capabilities. Content Operations performed by the Agent draw from that shared allowance, reducing the remaining balance as you use the Agent. The amount included depends on your plan. To purchase more Consumption Units, contact your Customer Success Manager or Account Manager.