Agent Bom, developed by Msaad00 for the mcp platform, is an open security scanner and self-hosted control plane. Built with Python, it supports AI, MCP, and cloud infrastructure. Distributed via PyPI, Docker, Helm charts, and Snowflake SPCS/Native App, it requires Python versions 3.11 through 3.14. This tool connects to and scans diverse systems, including repositories, software supply chains, identity, data infrastructure, and various cloud providers as read-only sources. It helps identify vulnerabilities and manage security posture across complex environments.

What does Agent Bom connect to?

Agent Bom functions as an open security scanner and control plane for AI, MCP, and cloud infrastructure. It connects to and scans various systems, including repositories, software supply chains, identity, data infrastructure, AI agents, and other MCP servers. The tool also establishes read-only connections to major cloud providers such as AWS, Azure, GCP, and Snowflake. This connectivity enables it to gather security insights across a diverse technology landscape, supporting a unified view of an organization's digital assets and potential vulnerabilities.

Core features and tools

Agent Bom exposes various callable CLI tools. Key tools include agent-bom scan for local/CI scans, agent-bom connect <provider> for cloud source connections, and agent-bom db update for advisory database updates. It also offers agent-bom serve to run the control plane, agent-bom agents for AI agent discovery, and agent-bom check for CVE gating. Specialized tools like agent-bom image for container scans and agent-bom iac for Infrastructure as Code scans are available. The MCP server mode exposes 6 resources and 8 workflow prompts, with the number of MCP tools reported as 77, 81, or 84. An optional youcom_search tool can be enabled by setting YDC_API_KEY.

Who is it for?

Agent Bom is designed for Developers/AI engineers, AppSec/product security professionals, Cloud security teams, Information Technology personnel, and System Administrators. It addresses the needs of those managing security and compliance across modern, distributed infrastructure, including AI and cloud environments. Its capabilities support proactive vulnerability identification and attack path analysis. This is not the right fit if you require a tool that writes configuration changes directly to your cloud providers during its core scanning functions.

How to get started

The authentication model for core scanning functions emphasizes a read-only approach, never storing credential values. For connecting to cloud sources, the setup involves adding a read-only grant from the provider. For the optional youcom_search MCP tool, authentication requires setting the YDC_API_KEY environment variable. MCP server communication supports stdio, SSE, and HTTP transport protocols.

Use cases

• Local and CI Scanning: Run "agent-bom scan" to discover dependencies, secrets, Infrastructure as Code, agents, and MCP servers, identifying vulnerable packages before deployment.
• Connected Source Scanning: Add read-only connections to cloud platforms like AWS, Azure, GCP, and Snowflake via "agent-bom connect" to run or schedule scans for broader infrastructure visibility.
• Blast Radius Analysis and Remediation Planning: Identify attack paths from vulnerabilities to exposed credentials and tools, then generate remediation plans using "agent-bom blast-radius drilldown" and "agent-bom agents --remediate".

Agent Bom provides a security scanning and control plane solution for AI, MCP, and cloud infrastructure, emphasizing read-only operations for its core functions. It offers a collection of CLI tools and MCP server features to help various security and engineering teams manage their digital assets. The tool supports multiple deployment methods and integrates with popular cloud providers.