chojs23/lazyagent

Lazyagent is a monitoring tool for developers who use AI coding assistants like Claude Code, Codex, or OpenCode. The problem it addresses is visibility: when these AI agents are running on your machine, they make decisions, spawn sub-agents, call tools, and consume tokens, but it can be hard to see exactly what they are doing and whether they are on track. Lazyagent gives you a real-time view into all of this activity. It works by hooking into each AI runtime's event system. When you run the init command for your chosen tool, it registers listeners for events like session start, tool use, prompts, and completions. All of that activity is stored in a local SQLite database. Lazyagent then provides two ways to view this data: a terminal-based interface (called a TUI, similar to tools like lazygit or lazydocker) that shows projects, sessions, agent hierarchies, and individual events in side-by-side panes, and an optional read-only browser dashboard you can open on a local port. You can see the full tree of which agent spawned which sub-agents, filter events by type, search across event content, and check token usage breakdowns showing how many tokens were spent and how many were saved by cache reuse. You would use this during active AI-assisted development sessions when you want to audit what the agent is doing, diagnose why a run went wrong, or understand your token costs. It is written in Go and available via npm, Homebrew, or direct Go installation.