bmdhodl/agent47

What it does

Agent47 is an MCP server that provides runtime guardrails and incident read access for coding agents. It enables developers to monitor and manage AI coding agents through AgentGuard traces, alerts, usage statistics, operational costs, and budget health monitoring. The tool integrates with existing AI development workflows to ensure agents operate within defined safety parameters and budget constraints.

How to set up

Agent47 is deployed as an MCP (Model Context Protocol) server. Installation involves: 1) Cloning the repository from bmdhodl/agent47, 2) Installing dependencies via package manager, 3) Configuring AgentGuard connection credentials, 4) Running the MCP server, 5) Connecting it to your preferred AI platform (Claude, ChatGPT, or AI Studio). Detailed setup instructions are available in the GitHub repository.

Key features

• AgentGuard trace querying and analysis
• Real-time alert monitoring and incident access
• Usage and cost tracking per agent
• Budget health and spending analysis
• MCP protocol integration for seamless platform compatibility
• Runtime guardrails enforcement for coding agents