Skip to content

Quick Start

Get a working test suite in under 5 minutes.

Prerequisites

  • Python 3.11+
  • Node.js 18+ (for the agent-browser CLI)
  • An LLM endpoint (Claude via claude-code-proxy, or any OpenAI-compatible API)

Install

pip install aitester-bdd
npm i -g agent-browser

Configure LLM access

# Default: claude-code-proxy (runs alongside Claude Code)
export AITESTER_LLM_MODEL=cc/claude-opus-4-7
export OPENAI_BASE_URL=http://localhost:20128/v1
export OPENAI_API_KEY=placeholder

Author a suite

aitester author \
  --story "Open Wikipedia, verify the search input exists, search for 'BDD', verify the article heading" \
  --base-url https://en.wikipedia.org \
  --out wiki_smoke.robot

The agent will:

  1. Open the browser and navigate to the base URL
  2. Take accessibility snapshots to ground selectors
  3. Write a .robot file with rules that test your story
  4. Dry-run it to verify the keywords parse

Run the authored suite

aitester run wiki_smoke.robot

Output: 

Wiki Smoke :: Wikipedia smoke test  | PASS |
1 test, 1 passed, 0 failed

No LLM tokens consumed at run time. The suite is plain Robot Framework.

Watch it run (headed mode)

aitester run wiki_smoke.robot --headed --step-delay 500

This opens a visible browser window and pauses 500ms after each action so you can follow along.

What just happened

  1. Author phase — the LLM drove agent-browser to explore the live site, took snapshots, grounded selectors in the real DOM, and composed a .robot file using the aitester-bdd keyword grammar.
  2. Run phase — Robot Framework parsed the .robot file. The aitester-bdd keyword library built an in-memory rule DAG (deferred execution). Then I finalize verification triggered the walker, which executed the DAG against a live browser.

Next steps