agentic-tester

Code-aware agentic tester for web apps.

1. Analyze your app (Next.js App Router first) into an app graph
2. Claude writes product flows + Action Specs (not Playwright)
3. Codegen compiles Action Specs → Playwright
4. Run headless
5. Audit when the app changes

Interactive CLI by default. Use -y for CI.

Install / run

npx agentic-tester init
npx agentic-tester view set mobile
npx agentic-tester storage set
npx agentic-tester generate
npx agentic-tester run
npx agentic-tester audit

Viewport (mobile vs laptop)

Mobile-only apps should run in phone view:

npx agentic-tester view set          # interactive
npx agentic-tester view set mobile   # iPhone 14 by default
npx agentic-tester view set laptop   # 1280×720 desktop
npx agentic-tester view show

This updates .agentic-tester/config.json and regenerates playwright.config.ts.

Or install locally:

npm i -D agentic-tester playwright
npx playwright install

Bring your own keys

ANTHROPIC_API_KEY=...
CLOUDINARY_CLOUD_NAME=...   # optional
CLOUDINARY_API_KEY=...      # optional
CLOUDINARY_API_SECRET=...   # optional

• Claude → writes/updates product flows + action specs
• Cloudinary → optional upload of run videos
• Without Claude key, heuristic flows are used from the app graph

localStorage JSON

File: .agentic-tester/localstorage.json

{
  "name": "daniel",
  "auth": {
    "token": "",
    "refresh": ""
  }
}

Translation:

| Key | localStorage value | |---|---| | name | "daniel" | | auth | "{\"token\":\"\",\"refresh\":\"\"}" |

Rules:

• top-level keys become localStorage keys
• strings stored as-is
• objects/arrays/numbers/booleans are JSON.stringify'd

npx agentic-tester storage set
npx agentic-tester storage show

Pipeline

Analyzer → app-graph.json
Claude/heuristic → flows/ + specs/
Codegen → generated/*.spec.ts
Runner → reports/ (+ optional Cloudinary links)
Audit → diff graph → update specs → regen

Source of truth: Action Specs in .agentic-tester/specs/
Generated output: Playwright in .agentic-tester/generated/

Commands

| Command | Purpose | |---|---| | init | Detect framework, analyze app, write flows/specs | | storage set | Interactive localStorage wizard | | generate | Compile action specs → Playwright | | run [flow] | Run all or one flow headless | | audit | Re-scan, diff, update, regenerate |

Requirements

• Node 18+
• Next.js App Router recommended for analysis
• Playwright installed in the target project to run tests

License

MIT