Turbocharge your testing with AI. Tresto pairs Playwright codegen with an agent that understands your intent and iterates toward robust, stable tests.

Convert manual checks into reliable automated E2E in minutes—no boilerplate, no lock‑in, just tresto.yaml at your project root.

• Manual testing everything is slow and fragile. It’s easy to miss regressions, hard to repeat precisely, and burns time every release.
• Writing E2E tests by hand is tedious. Locators break, timing is flaky, and keeping tests readable and maintainable takes effort.
• Tresto gives you the best of both worlds. Describe intent like you do in manual testing, and let AI produce durable, maintainable code you’d be proud to commit.

• Generates fully valid pytest + Playwright tests. No bespoke runner, no lock‑in. You keep industry‑standard tools and best practices.
• You stay in control. Ask the model to improve selectors, assert more precisely, or refactor flows—Tresto listens and iterates.
• Post‑release sanity. Tests started failing after your last release? Ask Tresto to investigate each failing test and determine if code needs updating or if it’s a real product bug.

• 🎯 Smart test generation: Natural-language to runnable Playwright tests
• 🎭 Playwright integration: Uses the Playwright Python stack
• 🤖 Agentic workflow: Generate → run → analyze → iterate
• 🧠 Multi-provider AI: Anthropic Claude and others via connectors
• ⚙️ YAML config: Single tresto.yaml at your project root
• 🧪 Pytest-native: Tests are discoverable and runnable with pytest

pip install tresto-ai

Or from source:

git clone https://github.com/LeaveMyYard/Tresto.git
cd Tresto
uv tool install --no-cache .

tresto init

This will create:

• test scaffold in your chosen directory (default: ./tresto/tests)
• a tresto.yaml configuration file

Then install Playwright browsers once per machine:

playwright install

# Open interactive AI-driven flow to create a test
tresto test create --test-name login.success   # optional name

# Iterate on an existing test with the agent
tresto test iterate --test-name login.success

# Run tests
tresto test run

• Python 3.11+ (3.13 preferred)
• Playwright browsers (playwright install)
• API key(s) for your selected AI provider(s)
  • For Anthropic set ANTHROPIC_API_KEY

After tresto init, edit tresto.yaml:

project:
  name: my-awesome-app
  url: http://localhost:3000
  test_directory: ./tresto/tests

ai:
  connector: anthropic
  model: claude-3-5-sonnet-20241022
  max_iterations: 5
  temperature: 0.1

browser:
  headless: true
  timeout: 30000
  viewport:
    width: 1280
    height: 720

recording:
  auto_wait: true
  capture_screenshots: true
  generate_selectors: auto

secrets:
  - ANTHROPIC_API_KEY

Notes:

• secrets is a list of environment variable names. They must be present in your environment; they are validated at startup.
• connector and model must be one of the values exposed by tresto models list.

• tresto: Shows a welcome panel and quick tips
• tresto init: Interactive setup; creates tresto.yaml and scaffolds tests
  • Options: --force, --template <name>
• tresto models list: List available AI connectors and their models
• tresto test: Alias for running tests (equivalent to tresto test run)
• tresto test run [PYTEST_ARGS...]: Run tests via pytest, forwards extra args
• tresto test create [--test-name <name>]: Start agent to create a test
• tresto test iterate [--test-name <name>]: Iterate on a test with the agent
• tresto db list-tests|show|clear|info: Inspect and manage test data storage
• tresto version: Show Tresto version

Deprecated/changed:

• .trestorc → replaced with tresto.yaml
• tresto record → use tresto test create and tresto test iterate

1. Inspect project and prompts based on your intent
2. Generate Playwright tests with the selected model
3. Run with pytest; collect logs, screenshots, insights
4. Iterate until assertions and flows are stable

Tresto is built on proven, open technologies:

• Python
• LangChain and LangGraph for agentic orchestration
• Playwright and Playwright codegen for robust, modern browser automation

Much thanks to the creators and maintainers of these projects—we stand on your shoulders.

• Automatic locator improvements across your codebase
• Improved processing of larger tests
• Supervisor agent that reviews the main agent’s resulting test
• Cloud model runner: access multiple providers from one subscription
• …and more improvements coming

See docs/CONTRIBUTING.md.

git clone https://github.com/LeaveMyYard/Tresto.git
cd Tresto
uv sync --dev
pre-commit install
pytest
ruff check .
mypy .

MIT — see LICENSE.

• Docs: ./docs/
• Issues: https://github.com/LeaveMyYard/Tresto/issues
• Discussions: https://github.com/LeaveMyYard/Tresto/discussions