# msaad00/agent-bom [Health: Active]

**Category:** 🔒 Security  
**Repository:** https://github.com/msaad00/agent-bom  
**GitHub Stars:** 36  
**Views:** 2  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/msaad00-agent-bom

## Description
AI supply chain security scanner with 18 MCP tools. Auto-discovers 20 MCP clients, scans dependencies for CVEs (OSV/NVD/EPSS/CISA KEV), maps blast radius from vulnerabilities to exposed credentials and tools, runs CIS benchmarks, generates CycloneDX/SPDX SBOMs, and enforces compliance across OWASP LLM Top 10, MITRE ATLAS, NIST AI RMF, and EU AI Act.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "agent-bom": {
    "command": "uvx",
    "args": ["agent-bom"]
  }
}
```

## Documentation

## What msaad00/agent-bom does

msaad00/agent-bom provides an AI supply-chain inventory and security scanner for agents, MCP servers, tools, packages, models, credential references, and connected resources. It builds relationships between these components so a dependency finding can be evaluated in the context of the agent and tools that may reach it.

Dependency analysis covers lockfiles and manifests across ecosystems including Python, JavaScript, Go, Java, .NET, and Rust. Findings are matched against OSV, GitHub advisories, and NVD, with EPSS and CISA KEV data available for prioritization. The scanner also supports native analysis of AI-related code, containers, infrastructure-as-code, cloud environments, and selected AI platforms.

Results can be mapped to OWASP LLM Top 10, OWASP Agentic Top 10, OWASP MCP Top 10, MITRE ATLAS, NIST AI RMF, and the EU AI Act. These mappings provide investigation context rather than certification.

## How it works

A local scan can inspect a repository and its MCP configuration, or skip workstation-wide discovery with `--no-discover`. Without a project configuration, package findings can still be produced, while agent and tool relationships require configuration or an inventory. The scanner records the collected environment as a graph linking agents, MCP servers, tools, packages, credentials, and data or cloud resources.

The graph represents a collected snapshot. Configured or modeled access links are not runtime observations, and recorded relationships do not by themselves establish exploitation or data access. Coverage should therefore be reviewed alongside findings before treating a scan as a complete assessment.

The project also provides an MCP proxy and gateway that can alert on poisoned tools and injected results, or block them when injection enforcement is enabled with `--injection-mode enforce`.

## Setup and configuration

For a repository scan, install `uv` and run:

```bash
uvx agent-bom scan . --no-discover -f html -o agent-bom-report.html
```

Open the generated HTML file for inventory, findings, coverage, and remediation guidance. A bundled offline example is available with:

```bash
uvx agent-bom scan --demo --offline
```

The project supports Python 3.11 through 3.14. Local scans do not require an account. Use `uvx agent-bom doctor` to investigate setup problems. The CLI can also be run in CI, and a GitHub Actions integration can upload SARIF and fail on high or critical findings using the default security gate.

The MCP interface is started through the `agent-bom mcp server` command and offers tool profiles for agent use. Claude Code and Codex plugins are also included.

## Tools and capabilities

- Discover agents, MCP servers, tools, packages, credentials, models, and supported connected resources.
- Scan dependencies against OSV, GitHub advisories, and NVD, with KEV and EPSS context.
- Trace vulnerability paths to exposed credentials, tools, agents, and reachable data.
- Run CIS benchmarks and compliance-oriented checks.
- Generate native agent-bom JSON inventories and CycloneDX or SPDX software BOMs.
- Export results as HTML, JSON, SARIF, CycloneDX, and SPDX.
- Import SARIF findings from other analysis tools.
- Inspect containers, Dockerfiles, Terraform, Kubernetes, Helm, and Compose configurations.
- Collect read-only posture and graph context from AWS, Azure, GCP, Kubernetes, Databricks, and Snowflake.

## Limitations and notes

The scanner’s graph describes the information collected for a scan rather than live execution behavior. Credential references are represented as names, not secret values. A non-zero exit can indicate either a security gate or an incomplete assessment, so the report and coverage should be inspected.

The project is self-hostable and licensed under Apache-2.0. Its control-plane option uses PostgreSQL, SSO/OIDC, roles, tenant isolation, audit records, and evidence import on the operator’s infrastructure. There is no managed cloud offering described; local, CLI, CI, and self-hosted deployment paths are available.

_Full upstream README: https://allmcps.com/mcp/msaad00-agent-bom/readme_