cleverpixels.net

What It Is

CogniWatch describes itself as an “AI Agent Security Observatory.” Its core goal is to continuously scan autonomous AI Agents across the internet, identify frameworks such as AutoGen, CrewAI, LangChain, and n8n, and surface them before they become security risks. Its positioning is not traditional known-threat detection, but rather coverage of an AI Agent infrastructure gap that is not yet well monitored.

Core Capabilities

Based on the page content, the platform includes modules such as Dashboard, Sentinel, Topology, Agents, MCP, and Admin. Sentinel handles continuous security posture assessment, showing threat scores, scan coverage, the number of monitored Agents, known vulnerabilities, and CVE checks. The Topology module visualizes global geographic distribution and port distribution; the Agents module provides a full inventory with search, filtering, and CSV export. The MCP module is the highlight: it can discover MCP Servers, identify exposed tool capabilities, classify risks, and run multi-turn adversarial testing based on Microsoft PyRIT, covering scenarios such as tool poisoning, confused deputy, trust laundering, and cross-origin escalation.

Pricing and Deployment

The page does not disclose pricing models, plans, trials, payment methods, or commercial support information. It also does not clarify whether the product is SaaS, privately deployable, or open-source self-hosted. The only things that can be confirmed are the presence of a protected admin panel, scan controls, and data export features. Details on compliance certifications, data retention, and permission management are also missing.

Pros and Cons

The main advantage is its forward-looking focus: it targets the exposed surface of AI Agents and MCP, covering discovery, fingerprinting, risk grading, vulnerability checks, topology visualization, and red-team validation. It is well suited for researching emerging AI infrastructure risks. The downside is that many page states are shown as Loading, scanning, or idle, making it hard to judge actual data quality. Enterprise procurement requirements such as SLA, compliance, audit trails, integration APIs, alerting channels, and support systems are not demonstrated.

Who It’s For and Access from China

It is better suited to security research teams, attack surface management teams, and AI platform security leads for mapping public-facing AI Agent/MCP exposure and producing risk reports. There is no information about access from China, payment, or localized support, so china_access can only be rated as unknown. For deployment in China, it may need to be complemented with local asset mapping, vulnerability management, SIEM/SOAR, or AI application security testing tools.