Show HN: Agent Chaperone – Screen AI agent tool calls and results with Jev

Category: security

Tags: ai-safety, agent-security, runtime-firewall, mcp, prompt-injection

Score: 7.3/10 (Innovation: 7, Technical: 6, Documentation: 9, Utility: 7)

Agent Chaperone is a runtime firewall for AI coding agents that screens tool calls before execution and tool results before the agent reads them, using a policy-driven approach with local logging. It stands out for its measured results, transparent limitations, and coverage of both MCP servers and client-specific hooks, making it a practical safety layer for agentic workflows.

Target audience: AI developers, platform engineers, DevOps, and security engineers deploying autonomous agents

Repository: https://github.com/agent-chaperone/agent-chaperone · TypeScript · Apache-2.0 · 23 stars

View on Hacker News