SecAIQ Watch

Product Introduction

1. Definition: SecAIQ Watch is a free, open-source, local-first AI observability and security dashboard. Technically, it is a PHP and SQLite-based application that functions as a passive monitoring agent and web-based control panel for AI-powered tools running on a user's computer.
2. Core Value Proposition: It exists to provide complete visibility and governance over local AI tool activity without compromising privacy. Its primary value is enabling developers, security professionals, and privacy-conscious users to see which AI assistants, coding agents (like Claude Code and Cursor), local LLMs (via Ollama, LM Studio), and MCP servers are active, what they connect to, what system permissions they hold, and what resources they consume—all from a local, read-only dashboard with no cloud dependency.

Main Features

1. Comprehensive AI Tool Detection: SecAIQ Watch uses signature-based detection to identify approximately 37 different AI tools and frameworks, including Claude Desktop, GitHub Copilot, Cursor, Windsurf, Codeium, Gemini, Aider, and various MCP servers. It also recognizes connections to ~33 common AI provider domains (e.g., api.openai.com, api.anthropic.com). Detection works by parsing system process lists and network connections on macOS, Linux, and Windows.
2. Real-time Network Activity Monitoring: The tool provides a live view of outbound connections from AI processes, displaying destination IPs/domains, ports, and the volume of data (bytes sent/received) per tool. It includes anomaly detection algorithms to flag unusual upload spikes or baseline deviations, offering critical insight into potential data exfiltration.
3. System Permission & Risk Analysis: This feature scans and categorizes the filesystem and system permissions accessible to each AI tool. It maps access to sensitive areas like SSH keys, .env files, cloud credential stores, browser data directories, and the system keychain. For each identified risk, it provides platform-specific (macOS, Linux, Windows) "how to remove" guides and one-click "Protect" presets to harden configurations for tools like Claude Code.
4. AI-BOM (Bill of Materials) & Reporting: A key innovation is the generation of an AI-BOM in CycloneDX 1.5 format, a standardized software bill of materials tailored for AI components. This, alongside HTML, Markdown, CSV, and JSON export capabilities, provides auditable evidence of AI tooling posture for compliance, security reviews, and asset management.

Problems Solved

1. Pain Point: The "black box" nature of local AI tools. Users often have no visibility into what background AI processes are doing, where they are sending data, or what sensitive files they can potentially access, creating significant security and privacy blind spots.
2. Target Audience: The primary personas are Security Engineers and AppSec teams needing to audit and govern AI tool usage within developer environments; Software Developers and DevOps Engineers who use AI coding assistants and want to understand their resource usage and security footprint; and Privacy-Conscious Professionals in regulated industries who must ensure local AI tools comply with data handling policies.
3. Use Cases: Essential scenarios include conducting a security audit before deploying a new AI coding agent across a development team; monitoring for unexpected network connections from a local LLM service; generating an asset report (AI-BOM) for compliance frameworks; and understanding the cost implications of token usage by different AI models and projects.

Unique Advantages

1. Differentiation: Unlike cloud-based SaaS monitoring tools or endpoint detection and response (EDR) systems, SecAIQ Watch is entirely local, read-only, and purpose-built for AI tools. It requires no account, sends no data to the cloud, and does not block or interfere with traffic, setting it apart from both traditional security software and generic network monitors.
2. Key Innovation: Its integration of passive, cross-platform observability with actionable, tool-specific hardening guides. The combination of deep technical detection (process, network, file handles) with contextual, step-by-step remediation advice for specific AI applications (like disabling a risky setting in Claude Code) is a unique approach to AI tool governance.

Frequently Asked Questions (FAQ)

1. Is SecAIQ Watch a firewall or does it block AI tool connections? No, SecAIQ Watch is a strictly read-only monitoring and observability tool. It is designed to provide visibility and awareness but does not block, filter, or modify any network traffic or file access. Its "Protect" features work by guiding users to change the AI tool's own configuration settings.
2. Does SecAIQ Watch work on Windows for complete monitoring? SecAIQ Watch for Windows is in beta and currently provides process and connection monitoring, but lacks per-connection byte counters and the open-file view available on macOS and Linux. Windows users can monitor which AI tools are running and what they connect to, but detailed network throughput and file handle data are limited.
3. How does SecAIQ Watch ensure its own security and not become an attack vector? The application enforces multiple security controls: it only accepts connections from the local machine (loopback interface), validates a secret token and same-origin headers for all actions, stores its database and logs with strict file permissions, and implements a rigorous Content-Security-Policy. A dedicated security-check.sh script is provided to audit a live installation.
4. Can SecAIQ Watch monitor AI tools that use local LLMs like Ollama? Yes, SecAIQ Watch specifically detects local LLM servers and frameworks such as Ollama and LM Studio. It can identify their processes and monitor any network connections they make, which is useful for understanding if a local model is attempting to call out to external services unexpectedly.
5. What is an AI-BOM and why is it important? An AI-BOM (AI Bill of Materials) is a machine-readable inventory of all AI components, models, and tools used in a system. SecAIQ Watch generates this in the standardized CycloneDX format, which is crucial for security audits, compliance (e.g., proving no unauthorized AI tools are in use), and software supply chain transparency in enterprise environments.