ShadowProtect: Wireshark for AI Agents

priyanshhhhh2 pts0 comments

shadowprotect · PyPI

Skip to main content<br>Switch to mobile version

Warning

You are using an unsupported browser, upgrade to a newer version.

Warning

Some features may not work without JavaScript. Please try enabling it if you encounter problems.

Search PyPI

Search

shadowprotect 0.1.0

pip install shadowprotect==0.1.0

Copy PIP instructions

Latest release

Released:<br>Jun 1, 2026

Wireshark for AI Agents: Real-time intrusion detection and protection.

Navigation

Verified details

These details have been verified by PyPI<br>Maintainers

Sensei_pri.21-12-2001

Unverified details

These details have not been verified by PyPI<br>Project links

Bug Tracker

Homepage

Meta

License: MIT License

Author: Your Name

Requires: Python >=3.11

Classifiers

License

OSI Approved :: MIT License

Operating System

OS Independent

Programming Language

Python :: 3

Topic

Security

Report project as malware

Project description

███████╗██╗ ██╗ █████╗ ██████╗ ██████╗ ██╗ ██╗<br>██╔════╝██║ ██║██╔══██╗██╔══██╗██╔═══██╗██║ ██║<br>███████╗███████║███████║██║ ██║██║ ██║██║ █╗ ██║<br>╚════██║██╔══██║██╔══██║██║ ██║██║ ██║██║███╗██║<br>███████║██║ ██║██║ ██║██████╔╝╚██████╔╝╚███╔███╔╝<br>╚══════╝╚═╝ ╚═╝╚═╝ ╚═╝╚═════╝ ╚═════╝ ╚══╝╚══╝

██████╗ ██████╗ ██████╗ ████████╗███████╗ ██████╗████████╗<br>██╔══██╗██╔══██╗██╔═══██╗╚══██╔══╝██╔════╝██╔════╝╚══██╔══╝<br>██████╔╝██████╔╝██║ ██║ ██║ █████╗ ██║ ██║<br>██╔═══╝ ██╔══██╗██║ ██║ ██║ ██╔══╝ ██║ ██║<br>██║ ██║ ██║╚██████╔╝ ██║ ███████╗╚██████╗ ██║<br>╚═╝ ╚═╝ ╚═╝ ╚═════╝ ╚═╝ ╚══════╝ ╚═════╝ ╚═╝

ShadowProtect — Wireshark for AI Agents

Real-time intrusion detection AND active protection for multi-agent AI systems.

The Problem

Multi-agent AI systems are the next major attack surface — and nobody is watching.

Prompt injection is invisible. An attacker poisons one agent's input; that agent silently propagates the instruction to every downstream agent in the pipeline. By the time you notice, the entire system is compromised.

No observability. Current agent frameworks expose no equivalent of network packet capture. You cannot see what agents say to each other, which tools they call, or how fast they communicate.

Zero protection. Every existing security tool for AI only detects attacks after the fact. ShadowProtect intercepts and blocks them before they execute.

The Solution

ShadowProtect wraps your existing agents with a 9-layer detection engine and a 12-module active protection layer — with one line of code, no refactoring required.

┌──────────────────────────────────────────────────────────────────┐<br>│ Your Agents (CrewAI / OpenAI / LangChain / any callable) │<br>│ ┌────────────┐ ┌────────────┐ ┌────────────┐ │<br>│ │ Researcher │──▶│ Planner │──▶│ Executor │ │<br>│ └────────────┘ └────────────┘ └────────────┘ │<br>│ │ │ │ │<br>│ ───────┴───────────────┴────────────────┴────────── │<br>│ shadowprotect SDK (pip install shadowprotect) │<br>│ ─────────────────────────────────────────────────────── │<br>│ Input Sanitizer → Scope Enforcer → Rate Limiter → Quarantine │<br>│ ───────────────────────────────────────────────────── │<br>│ FastAPI Backend │ 9 Detection Layers │<br>│ ───────────────────────────────────────────────────── │<br>│ Real-Time Dashboard (Next.js 14) │<br>└──────────────────────────────────────────────────────────────────┘

Installation

SDK (for your agents)

Install the lightweight Python client from PyPI:

pip install shadowprotect

Requirements: Python 3.11+

Backend + Dashboard (self-hosted)

The backend processes events, runs detectors, and serves the real-time dashboard. Run it alongside your agents:

Option 1: Docker (recommended)

git clone https://github.com/your-org/shadowprotect.git<br>cd shadowprotect<br>docker-compose up --build

Open http://localhost:3000 — dashboard is live.

Option 2: Manual

# Backend<br>cd backend<br>python -m venv .venv && .venv\Scripts\activate # Windows<br>pip install -r requirements.txt<br>uvicorn main:app --reload --port 8000

# Frontend (new terminal)<br>cd frontend<br>pnpm install<br>pnpm dev

Quick Integration — One Line

from shadowprotect import monitor

# Wrap any agent — zero code changes to your agent<br>monitored_agent = monitor(your_agent, backend_url="http://localhost:8000")

# Use it exactly as before — transparent proxy<br>result = await monitored_agent.execute(task)

Framework Examples

CrewAI

from crewai import Agent<br>from shadowprotect import monitor

researcher = Agent(role="Researcher", goal="...", backstory="...")<br>researcher = monitor(researcher, backend_url="http://localhost:8000")

# Use it exactly as a normal CrewAI Agent

OpenAI Agents SDK

from agents import Agent<br>from shadowprotect import monitor

agent = Agent(name="Planner", instructions="...")<br>agent = monitor(agent, backend_url="http://localhost:8000")

LangChain

from langchain.chains import LLMChain<br>from shadowprotect import monitor

chain = LLMChain(llm=llm, prompt=prompt)<br>chain = monitor(chain, backend_url="http://localhost:8000")

# Patches .invoke(), .run(), .arun(), .ainvoke() automatically<br>result = await chain.invoke({"input":...

shadowprotect agent agents from monitor import

Related Articles