Runs

A run is a single execution of a data extraction against a website. Each run opens a browser, navigates the page, and attempts to return data matching your schema.

Run lifecycle

Queued — the run is waiting for a browser session to become available.

Running — the agent or playbook is actively executing. The live view is active.

Completed — data has been extracted and validated against the schema successfully.

Failed — something went wrong. Common causes: the page changed, the login session expired, a script error, or a timeout. The error message and any partial results are preserved.

Agent types

Ev3ry uses two agent implementations depending on how the run is configured:

Single-page agent

The default. The agent explores the target page thoroughly:

1. Reads the DOM
2. Intercepts network requests made by the page
3. Checks framework globals (window.__NUXT__, __NEXT_DATA__, etc.)
4. Chooses the most reliable data source
5. Writes an extraction script
6. Validates output against the schema — retries if needed (up to 12 attempts)

This agent produces the most accurate results and handles API-rich SPAs, SSR frameworks, and complex JavaScript-heavy pages.

Multi-page agent

Used when extracting from two structurally identical pages at once (e.g., two match detail pages). The agent reads the cleaned HTML from both pages and writes a single extraction script that works for both. It does not have a validation loop — it's a one-shot extraction designed for DOM-consistent pages.

Live view

While a run is active, the live view shows:

• A real-time view of the browser the agent is controlling
• A log of actions the agent is taking
• Data rows as they're collected

Local browser live view uses polling-based screenshots. Cloud browser live view uses an embedded iframe (real-time). Both show the same information; the difference is only in update latency.

Run results

After a run completes:

• View extracted data in a table
• Download as JSON or CSV
• Save the run as a playbook (for AI-driven runs)
• Compare results with previous runs

Run history

All runs are preserved with their full action log, extracted data, and any errors. Access run history from the website detail page.