agent-pulse-skills

This Skill addresses the lack of a unified monitoring tool for local AI Agent sessions, costs, and performance. Developers who regularly use multiple AI coding agents (such as Claude Code, Codex, Cursor, etc.) often struggle to track token consumption, model invocation costs, response latency, and health status for each session. Agent Pulse Skills provides a comprehensive command-line toolset, enabling users to quickly gain insights into the operational status of their local AI Agents.

Usage is straightforward. First, you need to install the Python package agentpulse-cli separately. Then, install this Skill package via the npx skills add command. After that, you can use natural language instructions to perform various monitoring operations, such as viewing session summaries from the last 24 hours, querying the most costly models, comparing activity across different platforms, troubleshooting issues like a dashboard showing no data, or exporting reports and MCP tool lists. The system includes a rich set of subcommands like status, top, session, models, health, and forecast, covering needs from daily overviews to in-depth analysis.

It is particularly suitable for teams or individuals who frequently use multiple AI coding agents, especially those who have deployed local models (such as Hermes, DeepSeek, Claude, Codex, etc.) and wish to precisely control costs and performance. DevOps engineers, backend developers, and data analysts can leverage this tool to easily grasp AI Agent resource consumption trends, optimize model selection, and adjust strategies before exceeding budgets.

It is recommended to run agent-pulse status once before starting a new AI Agent session to establish a baseline. Note that this Skill itself does not include the Agent Pulse runtime; you must first install the CLI via pip install agentpulse-cli and ensure a functional Python environment. On Windows systems, you need to set the UTF-8 environment variable to prevent garbled characters during command execution.