Skip to main content

Agentic Programming with Boozang

Give this URL to your AI assistant

https://docs.boozang.com/agent/AGENT.md — your AI can read this directly for full Boozang context, concepts, workflows, and API reference.

Agentic programming refers to using AI assistants to help create, manage, and maintain automated tests. Boozang is designed from the ground up to support this paradigm.

Quick Start​

npx bz-agent init

That's it. This generates a .bz/ directory with everything your AI assistant needs — configuration, authentication, and complete documentation (concepts, workflows, best practices, and the full MCP API reference with 40+ tools).

Setup Steps​

  1. Generate an MCP token in Boozang: Settings -> API Tokens
  2. Run npx bz-agent init — enter your server URL and token
  3. Point your AI assistant to .bz/AGENT.md
  4. Describe your tests — the AI creates and manages them via the MCP API

What bz-agent init Creates​

.bz/
├── AGENT.md          # Entry point — give this to your AI
├── config.json       # Server URL, API endpoint, version
├── .env              # Auth token (gitignored)
├── mcp-api.md        # Full API reference (40+ tools with parameters)
├── concepts.md       # Data model and terminology
├── workflows.md      # Common automation patterns
└── best-practices.md # Test design recommendations

Other Commands​

npx bz-agent status     # Test connection and show config
npx bz-agent snapshot   # Cache project structure locally for AI
npx bz-agent update     # Re-render docs from latest templates
npx bz-agent parse <log> # Parse Boozang runner logs for failures

Why Boozang for Agentic Programming?​

Every Web Application is a State Machine​

Boozang's core insight: users don't script click sequences — they model their application's behavior. This maps naturally to how AI thinks about applications.

Natural Language Foundation​

Boozang identifies elements using natural language rather than brittle CSS selectors or XPaths:

  • Tests read like documentation
  • AI can understand test intent from the test itself
  • Changes to test descriptions map directly to UI changes

Structured Data Model​

Tests follow a clear hierarchy that AI can navigate programmatically:

Company → Project → Version → Module → Test → Action

MCP API (40+ Tools)​

The Model Context Protocol API gives AI agents full control:

  • Module/Test/Action CRUD — create, read, update, delete at every level
  • Environment management — create, provision, configure test environments
  • IDE control — navigate, run tests, get screenshots, set breakpoints
  • Auth configuration — manage authentication settings

Two Ways to Get Started​

ApproachBest forWhat to do
npx bz-agent initProject-specific setupRun in your project directory. AI reads local .bz/AGENT.md
Public URLQuick one-off helpGive your AI: https://docs.boozang.com/agent/AGENT.md

Both provide the same documentation. The CLI approach adds project-specific config and authentication.

Next Steps​