Customize the AI

Shape Test Companion to fit your team’s standards, repeatable tasks, and IDE preferences using rules, agents, and configuration.

Test Companion works out of the box. Auto mode routes your prompts to the right built-in capability without any setup. Customization makes Test Companion match the way you and your team work. Three layers of customization are available, and each addresses a distinct need.

However, every team’s testing environment is unique. Your organization may have its own frameworks, coding standards, naming conventions, folder structures, and recurring processes. Customization is essential for teaching Test Companion about these specifics, so it can produce outputs that align with your project from the very first attempt, instead of requiring corrections each time.

There are three layers of customization, and they work together:

Understand the customization options

Picking the wrong layer leads to either rigid or inconsistent results. Use the guidance below to choose.

• Use rules when you want Test Companion to follow the same instruction every time without being asked. Rules apply automatically. The user does not invoke them. A rule is the right choice when the instruction is short, applies to many different tasks, and should never be skipped.

• Use agents when you have a multi-step task you run more than once. Agents are invoked explicitly with /agent-name when you need them. An agent is the right choice when the task has a defined sequence of steps, takes context as input, and produces a predictable output.

• Use configuration and preferences when you need to change how the extension itself functions, not what it produces. Configuration controls behavior such as viewport size, approval prompts, and which BrowserStack account is in use.

The three layers compose. A workspace can have rules that always apply, agents that you invoke for specific tasks, and configuration that defines how Test Companion runs in your IDE. Together they let you tune the assistant without writing custom code.

Quick-start path

If you are setting up Test Companion for the first time and want the fastest path to useful results, do these three things in order.

1. Write one rule that captures your team’s most important standard.
   For example, Always use explicit waits in Playwright tests or Never modify files in the config/ directory. Add it as a global rule, so it applies everywhere. Refer to Rules documentation for the full procedure.

2. Create one workspace agent for a task you run more than once a week. A pre-merge regression check, a deploy script, or a release-notes generator are good first agents. Refer to Agents documentation for the file format and invocation steps.

3. Review your interaction mode in Configuration and preferences. Decide whether you want Test Companion to ask for approval before editing files, or to act autonomously within boundaries you set. The default is conservative. Adjust as you build trust.

Next steps

• Rules: Learn how to create persistent, background guidance that shapes the AI’s behavior across all tasks.
• Configuration and preferences: Explore the full Settings panel, interaction modes, system instructions, auto-approve controls, viewport options, and more.