AperionAI/shield

Overview

Local guardrails for AI coding agents. Wraps any MCP server and blocks destructive tool calls like DROP TABLE, rm -rf, force-push, and unscoped UPDATE/DELETE before they execute. Free, open-source, runs entirely on your machine.

Best for

Best for
Developers using AI coding agents who need to prevent accidental destructive actions on their local machines.

Use cases

• Preventing accidental database drops during AI-assisted code generation
• Blocking destructive file system commands from AI agents in development environments
• Enforcing safe git operations by restricting force-push from AI tools

How to use

Tools exposed

• RULE_NEVER_FIRES
• CONSISTENTLY_DEMOTED

Tested with

Cursor, Claude Code, Cline, Continue, Windsurf, Zed

Notes

Local guardrails for AI coding agents. Wraps any MCP server and blocks destructive tool calls like DROP TABLE, rm -rf, force-push, and unscoped UPDATE/DELETE before they execute. Free, open-source, runs entirely on your machine.

5 stars on GitHub. Last updated 2026-07-03. Licensed Apache-2.0.

Use cases

• Preventing accidental database drops during AI-assisted code generation
• Blocking destructive file system commands from AI agents in development environments
• Enforcing safe git operations by restricting force-push from AI tools

Pros

• Free and open-source with no vendor lock-in
• Runs locally, preserving data privacy and reducing latency
• Works with any MCP server, making integration straightforward

Cons

• Only applicable to tools using the MCP protocol, not standalone agents
• May block legitimate commands if rules are too conservative, requiring tuning
• Requires manual setup to whitelist safe operations

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.