e2e-base
Where the other ddakit bases are starting points for building an app, e2e-base is a starting point for automatically verifying that an app behaves the way it’s described. The target is a separate external app you already have (the SUT), and the output is a test suite that verifies it. The AI harness analyzes the target app and picks the tools: Playwright for web, Playwright _electron for electron, Maestro and Detox for mobile.
This page is just the fork. The single table below picks your row, then you jump straight to that spot.
Where to start
Section titled “Where to start”| If you are | Go to |
|---|---|
| Not a full-time developer — first time with a terminal, with git, with testing tools | Beginner manual |
| Comfortable with React and TS, first time with an e2e tool like Playwright or Maestro | Beginner manual, prerequisites through first-run, then Expert manual |
| Already built an e2e suite with Playwright or Maestro | Expert manual, straight to architecture |
If none of the three rows is an obvious fit, pick the first one. The beginner manual moves one action per step, so anything that already feels familiar you can skim past.
How the two manuals differ
Section titled “How the two manuals differ”The expert manual reads as short reference. It assumes the words POM, fixtures, storageState, sharding, and ADR already feel familiar, and it stays close to the reasoning behind each decision. Seven sections — intro, architecture, tracks, analyze-and-design, verification, ci, and updates.
The beginner manual moves step by step. It assumes you are comfortable with the idea of an AI writing test code, but maybe new to a terminal, to git, to Node, or to the testing tools. The text walks alongside you and tells you what the screen should look like when things are right. Index plus ten steps.
Two halves of the same base, told in two paces. The tone differs but the tools and commands stay the same.
What’s the same either way
Section titled “What’s the same either way”The first run only bootstraps the test environment. You don’t have to have picked the app to verify yet. You set up the environment with bash init.sh, choose your pace with /start in Claude Code, and then analyze the target and its critical flows once your app has a shape worth testing. The flow is cut into /1-pace through /6-cleanup-residue up front, with /run-suite and /review-e2e at the end. One step has to finish before the next one starts.
The verification target is local, staging, or a dedicated test environment only. A production app and a production database are never the target. Real keys live in .env.test.local and are never committed. Both the beginner and the expert manual repeat this boundary.
The AI entry points are /start (one line in Claude Code) and AGENTS.md at the repo root for Codex. Both routes point at the same canonical doc, AI_AUTOMATION.md.
The license is All Rights Reserved with a commercial license model on top. Payment flow and a lawyer-reviewed COMMERCIAL-LICENSE.md are still in progress, which means actual sales wait on those two. The manual and the landing are public, so this page you are reading stays public as it is.
Once you picked your row, head straight to the Expert manual or the Beginner manual.