MCP Agent Health Reporter

Scans a local directory of AI agents and returns AOS readiness scores, section breakdowns, and audit summaries before deployment.

What you can do

Provides two tools: aos_agent_health_scan scores a single agent under a target directory, returning an aos_score, a certified flag, per-section scores (manifest declaration, systemd runtime, immune loop, physical evidence), and precedent metadata; aos_agent_health_audit scans a directory and returns totals such as total_tools and avg_score with per-agent reports. It is advisory, intended to catch agents that look healthy locally but fail once deployed.

When to use it

Use it when you keep a directory of agent definitions and want a pre-deployment readiness check or a batch audit of several agents. It suits local development workflows where a scored report is more useful than manual inspection.

Requirements

Runs locally over stdio as the PyPI package mcp-agent-health, launched with uvx or installed via pip. Python is required. Optional environment variables: AOS_HEALTH_TARGET_DIR sets the default scan root (fallback is the current working directory), AOS_HEALTH_MCP_LOG sets the JSONL log path, and AOS_HEALTH_MCP_TRANSPORT selects the transport (stdio by default). No accounts or API keys are declared.