A developer released "What's Agent Doing," an open-source plugin for Claude Code that surfaces what the agent is doing mid-run. Built on Claude Code's function hooks, it displays the current step, timers, and background-agent rows above the prompt. It addresses the opacity of newer coding models running long, obscure command sequences.
A tool that helps developers using terminal-based AI coding agents (like Claude Code) keep track of changes made by the agent. It provides a file viewer with diffs to see what actually changed, and a view of the agent's actions, solving the problem of losing context when the agent iterates on code while the developer zones out.
A tool that helps developers using terminal-based AI coding agents (like Claude Code) keep track of changes made by the agent. It provides a file viewer with diffs to see what actually changed, and a view of the agent's actions, solving the problem of losing context when the agent iterates on code while the developer zones out.
rich-statusline is a customization mod for Claude Code that renders a richer statusline under the prompt. It shows context usage split by category, 5-hour and weekly usage limits with reset countdowns, and project details like repo, branch, worktree, and PR. It targets Claude Code users who want more visibility and layout control than the built-in statusline.
rich-statusline is a customization mod for Claude Code that renders a richer statusline under the prompt. It shows context usage split by category, 5-hour and weekly usage limits with reset countdowns, and project details like repo, branch, worktree, and PR. It targets Claude Code users who want more visibility and layout control than the built-in statusline.
AgentGlass reads Claude Code transcripts on your machine and displays live what Claude is doing in each tmux session: tool calls, files touched, stuck loops, cost per session and per model. It solves the problem of not knowing what Claude Code is actually doing inside terminal panes, for developers who run multiple Claude Code sessions.
AgentGlass reads Claude Code transcripts on your machine and displays live what Claude is doing in each tmux session: tool calls, files touched, stuck loops, cost per session and per model. It solves the problem of not knowing what Claude Code is actually doing inside terminal panes, for developers who run multiple Claude Code sessions.
RayTrace is a sandboxing service for coding agents that records what Claude Code actually executes on a virtual machine. It targets developers who rely heavily on AI coding agents and want safe, observable execution. It is in a pre-pilot stage seeking five testers with $10 credit.
RayTrace is a sandboxing service for coding agents that records what Claude Code actually executes on a virtual machine. It targets developers who rely heavily on AI coding agents and want safe, observable execution. It is in a pre-pilot stage seeking five testers with $10 credit.
hud is an open-source terminal UI that shows a compact, always-visible status display for coding agents like Claude Code, Codex, and OpenCode. It solves the problem of terminal flooding by showing only key signals: running state, current action, and last message. It lets users monitor and interact with agents without reading the full log.
hud is an open-source terminal UI that shows a compact, always-visible status display for coding agents like Claude Code, Codex, and OpenCode. It solves the problem of terminal flooding by showing only key signals: running state, current action, and last message. It lets users monitor and interact with agents without reading the full log.