"Understanding Code Execution" matching MCP servers:

A security filter that blocks dangerous code patterns by comparing normalized structural syntax trees against a blacklist of known threats using vector embeddings. It acts as a gatekeeper to prevent malicious code execution by identifying dangerous structures regardless of specific identifiers or literals.

Biometric authorization for AI agent actions via Face ID on iPhone, enabling secure approval of sensitive actions and credential-safe API calls through vault execution.

Deterministic AI safety policy engine with Z3 formal verification. Write, verify, simulate, and enforce machine-verifiable safety constraints for AI agents. Completely outside the LLM.

A tool for detecting and cleaning Java memory shells via local or SSH remote execution. It enables AI agents to scan Java processes, analyze suspicious class code, and safely remove memory shells after user confirmation.

Provides a human-approval tool for MCP-compatible agents, pausing execution until a human approves or edits the action, with every decision logged.

Runtime policy enforcement for AI agents. Evaluate every agent action against your organization's policies before execution, with observe and enforce modes.

An MCP server that enforces plain-English rules on AI agent tool calls, deterministically blocking forbidden actions before execution via a Claude Code hook and a check_action tool.

Enables Claude Desktop and Claude Code to inspect a Postgres schema and run read-only SQL queries against a demo database through the tools describe_schema and query_readonly. Every proposed query passes the same deterministic guardrails — SELECT-only, allow-listed tables, enforced row limits, statement timeouts and a read-only role — before execution.

Provides real-time security risk assessment for AI coding prompts, analyzing potential dangers, blast radius, and complexity before code execution in Cursor.

A compact MCP server demonstrating explicit tool boundaries, least-privilege discovery, execution-time authorization, destructive-action confirmation, and metadata-only audit logs using a local note store.

Enables AI to read and analyze repository incident scan results, including file diffs and summaries, to help investigate suspicious changes. It is read-only and does not execute code or send data externally.

Governor is an MCP server that acts as a safety belt for AI agent payments, providing a three-layer deterministic evaluation (rules, behavior anomaly, and intention coherence) to approve, escalate, or block payments before execution, and integrates directly with Claude Code.

Enables forensic investigators to acquire data from Android devices via ADB, including device backup, artifact collection, and secure command execution.

Security gate that scans MCP servers and Claude Code Skills for vulnerabilities before execution.

MCP server for angr — symbolic execution + CFG + reaching-definitions.

Strict AI code reviewer powered by Groq. Finds bugs, vulnerabilities and security issues in your code. Supports analyze, compare versions, explain code, generate tests, and HTML reports.

Indexes a workspace and exposes code-navigation tools to Codex, enabling symbol lookup, reference finding, dependency analysis, and security audit for local codebases.

Enables safe GitHub automation by queuing write operations for admin approval before execution.

CodeSafer is an MCP server that scans AI-generated code for 9 categories of hidden security threats — including invisible Unicode, Trojan Source, homoglyphs, and rules file backdoors — using static analysis plus CodeBERT deep learning. Runs locally, free tier available.

Enables AI-driven open-source intelligence investigations by exposing 30 tools for scanning emails, usernames, domains, IPs, and more, with real subprocess execution.