agent-guardrail

ss1738/agent-guardrail

★ 1 stars Python Code/Dev Tools Updated today

A runtime gate for coding agents: an MCP server that blocks the tool calls that wreck a repo (force-push main, rm -rf, secret exfil, CI wipe) and lets real work through. Machine-checked git-branch core. Tested on 3,790 real commands.

View on GitHub → 🔍 Audit Wallet Slippage →

Quick Install

Copy the config for your editor. Some servers may need additional setup — check the README.

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "agent-guardrail": {
      "command": "uvx",
      "args": [
        "agent-guardrail"
      ]
    }
  }
}

Or install with pip: pip install agent-guardrail

README Excerpt

<p align="center"> <img src="docs/demo.svg" alt="A hijacked coding agent tries to force-push main, rm -rf the repo, leak a secret and wipe CI; agent-guardrail blocks every one while the real fix goes through" width="760"> </p> A runtime gate for autonomous coding agents. It runs as an [MCP](https://modelcontextprotocol.io) server in the tool-call path between the agent and your repo, so every `run_shell`, `write_file`, and `git` call is checked before it executes. It blocks the actions that wrec

Tools (1)

ToolGate

Topics

agent-safetyai-agentsai-safetycoding-agentscopilotformal-verificationguardrailsllmmcpmodel-context-protocolsecurityz3

agent-guardrail

Quick Install

README Excerpt

Tools (1)

Topics

Related Code/Dev Tools Servers