MON -- Moson Web Intelligence Engine

MON is a modular Python framework for inspecting websites: understanding frontend architecture, discovering backend APIs, generating API documentation, reconstructing application structure, and exporting structured knowledge -- e.g. for feeding into an AI agent.

It is not a website cloner. Cloning raw files is one small action among twelve; MON's real job is reconstruction: turning a live website back into a structured, documented, machine-readable specification of how it works.

Install

pip install -r requirements.txt
# or, for development (adds pytest):
pip install -e ".[dev]"

Usage

MON exposes exactly one public function.

from mon import inspect

result = inspect(
    domain="example.com",
    action="all_data",       # or e.g. ["explorer", "api_spec", "technology"]
    profile="balanced",      # "fast" | "balanced" | "deep"
    output_format="json",    # "python" | "json" | "markdown" | "html" | "yaml"
    output_dir="./output",
)

print(result.apis)           # list[Endpoint]
print(result.forms)          # list[Form]
print(result.routes)         # list[Route]
print(result.technology)     # list[Technology]
print(result.explorer_visual)

See examples/ for more: basic.py, explorer.py, api_spec.py, all_data.py.

Architecture

User -> inspect() -> SDK -> Inspector -> Resolver -> Dispatcher -> Analyzers -> Context -> Exporter -> InspectResult

• SDK (mon/sdk.py) -- the only public entry point. Builds config, calls Inspector, optionally saves output.
• Inspector (mon/engine/inspector.py) -- orchestrates the run. Knows nothing about HTML/JS/APIs.
• Resolver (mon/engine/resolver.py) -- expands composite actions (e.g. api_spec) and topologically sorts analyzers by declared dependencies, so javascript always runs before api, which always runs before relationships.
• Registry (mon/engine/registry.py) -- maps action names to analyzer classes. Adding a new analyzer never requires touching the Dispatcher.
• Dispatcher (mon/engine/dispatcher.py) -- executes the resolved pipeline against one shared InspectContext.
• Context (mon/engine/context.py) -- the only channel analyzers use to communicate. No analyzer ever imports another analyzer.
• Events (mon/engine/events.py) -- analyzers never call print(); they emit events, and ProgressManager (or a future plugin/AI layer) subscribes.
• Analyzers (mon/analyzers/) -- one class per file, one responsibility each: crawler, html, javascript, forms, api, routes, assets, metadata, technology, server, relationships, explorer.
• Exporters (mon/exporters/) -- turn an InspectResult into Python/JSON/Markdown/HTML/YAML.

Actions

Leaf actions: crawler, html, javascript, forms, api, routes, assets, metadata, technology, server_info, relationships, explorer.

Composite actions: all_data, frontend, backend, api_spec, explorer_visual, cloning.

Confidence scoring

Every Endpoint carries a confidence.score (0-100) and confidence.reasons -- a list of exactly why MON believes the endpoint is real (e.g. "Success/error branches detected in JS", "Matched to HTML form", "Live-verified against server"). Nothing is asserted without a reason.

License

MIT -- see LICENSE.