Decision Framework: When to Use What

Quick Decision Tree

Do you need browser automation?
├── No → Use standard tools (curl, API clients)
└── Yes
    ├── Is an AI agent driving it?
    │   ├── No → Use Playwright or Selenium directly
    │   └── Yes
    │       ├── Do you need semantic page understanding?
    │       │   ├── Yes → Playwright MCP (accessibility testing, exploration)
    │       │   └── No
    │       │       ├── Do you know the CSS selectors?
    │       │       │   ├── Yes → Vibium CLI skill (fastest, cheapest)
    │       │       │   └── No
    │       │       │       ├── Is the UI complex/custom? → browser-use (vision)
    │       │       │       └── Standard web UI → Vibium + find-all for discovery
    │       │       └── Is token budget critical?
    │       │           ├── Yes → Vibium CLI skill
    │       │           └── No → Any approach works
    │       └── Context window size?
    │           ├── <200K → CLI skill (mandatory for efficiency)
    │           └── >1M (Gemini) → MCP is acceptable
    └── What type of testing?
        ├── Functional → Vibium CLI skill
        ├── Accessibility → Playwright MCP
        ├── Visual regression → Vibium screenshots + comparison
        ├── Exploratory → Playwright MCP or browser-use
        ├── Performance → Traditional tools (k6, Lighthouse)
        └── API → No browser needed (curl, httpie)

Scenario-Based Recommendations

Scenario 1: New Startup, Building From Scratch

Context: Small team, modern SPA, CI/CD via GitHub Actions, Claude Code as coding assistant.

Recommendation: Vibium CLI skill

Why:

• Zero setup cost (one install command)
• Skills integrate naturally with Claude Code
• Token-efficient for a team that's also using the agent for code writing
• Good enough for a test suite of 50-200 tests
• Can add MCP later for specific needs

Scenario 2: Enterprise with Existing Selenium Suite

Context: Large team, legacy application, 2000+ existing Selenium tests, migrating to AI-assisted QA.

Recommendation: Selenium 4 (existing) + Vibium CLI skill (new tests)

Why:

• Don't rewrite 2000 tests — maintain existing Selenium suite
• New tests use Vibium skill for AI agent integration
• Both use WebDriver BiDi (converging protocols)
• Gradual migration as old tests naturally need updates

Scenario 3: Accessibility-First Product

Context: Government or healthcare app with WCAG compliance requirements.

Recommendation: Playwright MCP (primary) + Vibium CLI skill (functional tests)

Why:

• Playwright MCP's accessibility tree is essential for WCAG audits
• The agent can analyze ARIA roles, labels, and structure semantically
• Functional tests use CLI skill to save context for a11y analysis

Scenario 4: Complex B2B Dashboard

Context: Data-heavy application with charts, tables, drag-and-drop, custom components.

Recommendation: browser-use (complex interactions) + Vibium CLI skill (standard flows)

Why:

• Custom components may not have standard selectors
• Vision model can identify elements visually
• Standard flows (login, navigation) still use CLI for efficiency
• Higher cost is justified for complex UI

Scenario 5: CI-Only Test Runner (No Human Involvement)

Context: Fully automated pipeline, tests run on every push, no human reviews failures in real-time.

Recommendation: Vibium CLI skill (oneshot + headless)

Why:

• Lowest cost per run ($0.05/test)
• Fastest execution (no MCP overhead)
• Failure artifacts (screenshots + text) captured automatically
• JUnit XML for CI dashboard integration
• Agent's self-healing reduces false failures

Cost Comparison by Scale

Small Scale (50 tests, 5 runs/day)

Medium Scale (200 tests, 10 runs/day)

Large Scale (500 tests, 20 runs/day)

Costs assume $15/M input tokens (Claude Sonnet). Actual costs vary by model and provider.

The cost difference is dramatic at scale. This is why token efficiency isn't a nice-to-have — it's an economic necessity for production test automation.

Migration Paths

From Selenium to AI-Assisted

Phase 1: Add Vibium alongside Selenium
Phase 2: New tests use Vibium skill
Phase 3: Convert high-maintenance Selenium tests
Phase 4: Decommission old Selenium tests as they're replaced

From Playwright to AI-Assisted

Phase 1: Add Playwright MCP for exploration
Phase 2: Add Vibium skill for execution
Phase 3: Hybrid approach (MCP for discovery, skill for execution)
Phase 4: Most tests run via skill, MCP for accessibility only

From Manual QA to AI-Assisted

Phase 1: QA writes test definitions in YAML (natural language)
Phase 2: Agent executes definitions via Vibium skill
Phase 3: Agent generates new test cases from app exploration
Phase 4: Continuous autonomous QA (agent runs independently)

Risk Assessment

The Bottom Line

For your interview, the recommended stack is:

Primary: Vibium CLI skill for all functional test execution Secondary: Playwright MCP for accessibility audits and page discovery Framework: YAML test definitions + Bash test runner + GitHub Actions CI

This gives you:

• Cost efficiency for daily CI runs
• Semantic understanding when you need it
• Standards compliance (W3C BiDi)
• Self-healing through agent reasoning
• Simplicity (CLI commands, markdown skills, YAML tests)