How agents use data

Agents use your data through prompts, variables, and pipelines.

When an agent runs, Etals gives it the data it needs for the current item, market, language, brand, or website. The agent can then use that context to produce a new value for a property.

Reading data with variables

Prompts can include variables such as [product_name], [market], [language], or [product_description].

When the agent runs, Etals replaces those variables with the real values for the item being processed.

Write a meta description in [language] for the [market] market.
Product name: [product_name]
Current product description: [product_description]
Brand guidelines: [brand_guidelines]

This lets one prompt work across many products, markets, and languages.

Producing property values

Agents can also produce new values for properties.

For example, a copywriter agent can read supplier text and produce a product description. A translator agent can read that product description and produce a translated version. A validator agent can review the translated value before it is exported or published.

Using pipelines

Pipelines decide which agents run and when they run.

A common flow might look like this:

1. Read product data from Etals.
2. Generate a product description.
3. Translate the description into target languages.
4. Validate the translations.
5. Store the approved values back on the correct properties.

Keeping context specific

Agents usually perform better when they receive only the information they need.

Instead of putting every brand, market, and product rule into one large prompt, use prompt blocks and conditions to include the right context only when it applies.