Los mejores MCP servers de Security para Codex

Securely execute LLM-generated code in isolated containers or local Linux and Windows sandboxes across multiple programming languages.

[mcp_servers.llm-sandbox]
command = "uvx"
args = ["llm-sandbox"]

Enforces policy-as-code governance for AI agents through deterministic scheduling and verifiable audit receipts.

[mcp_servers.bernstein]
command = "uvx"
args = ["bernstein"]

Reverse engineer binaries, applications, and runtime behavior through an AI agent or terminal interface.

[mcp_servers.rea]
command = "npx"
args = ["-y", "rea-agents"]

Provides LLM agents with controlled, audited SSH access to remote Linux and Windows hosts using policy-based authorization and human-in-the-loop approvals.

[mcp_servers.ssh-mcp]
command = "npx"
args = ["-y", "ssh-mcp"]

Provides encrypted, fully offline, and durable agentic memory for AI agents to store and recall facts, preferences, and decisions locally.

[mcp_servers.compartment]
command = "uvx"
args = ["compartment"]

Scans and investigates AI supply chains by mapping agents, MCP servers, packages, and their potential reach to identify vulnerabilities.

[mcp_servers.agent-bom]
command = "uvx"
args = ["agent-bom"]

Provides an AI-powered OSINT toolkit that lets agents perform passive reconnaissance using tools for DNS, WHOIS, IP, and breach lookups.

[mcp_servers.openosint-2b2261]
command = "uvx"
args = ["openosint"]

Provides an AI-powered OSINT toolkit that lets agents perform passive reconnaissance using tools for DNS, WHOIS, IP, and breach lookups.

[mcp_servers.openosint]
command = "uvx"
args = ["openosint"]

Lets AI clients run private web searches, read URLs, and get query suggestions through a SearXNG instance.

[mcp_servers.mcp-searxng]
command = "npx"
args = ["-y", "mcp-searxng"]

Enables AI agents to perform security research, vulnerability analysis, and automated code fixing through a specialized security harness.

[mcp_servers.0]
command = "npx"
args = ["-y", "0sec-cli"]

Enables AI assistants to perform professional network analysis, security audits, and threat detection using tshark and packet capture files.

[mcp_servers.wireshark-mcp]
command = "uvx"
args = ["wireshark-mcp"]

Provides a chain of custody for AI agents by issuing signed receipts for tool calls and enforcing access policies via an MCP gateway.

[mcp_servers.agent-custody]
command = "npx"
args = ["-y", "@agent-custody/receipts"]

Validates architectural decision records against your actual code, detects drift, masks sensitive content, and tracks decisions.

[mcp_servers.mcp-adr-analysis-server]
command = "npx"
args = ["-y", "mcp-adr-analysis-server"]

Enables AI clients to securely access metadata management capabilities including search, lineage, data quality, and governance through OAuth 2.0.

[mcp_servers.openmetadata-mcp]
url = "https://{openmetadata_host}/mcp"

Hardens binaries, Electron apps, JS bundles, Python, and WebAssembly artifacts against AI reverse-engineering agents.

[mcp_servers.bla]
command = "npx"
args = ["-y", "bla-agents"]

Analyze customer feedback by stripping PII before processing to find stable themes and provide cited evidence for AI agents.

[mcp_servers.obsei]
command = "uvx"
args = ["obsei"]

Puts authorization in front of AI agent HTTP actions, forwarding only requests approved by the Decionis authority and holding or blocking the rest.

[mcp_servers.agent-safe-pipeline]
command = "npx"
args = ["-y", "@decionis/mcp"]

Allows AI clients to search a community library of threat hunting hypotheses by keyword, threat actor, or available log sources.

[mcp_servers.hearth]
command = "npx"
args = ["-y", "@thorcollective/hearth-mcp"]