All shelves / Mods / Workflow
agent-monitor
By GarvitNangru
agent-monitor is a Claude Code workflow mod by GarvitNangru.
Draws a progress bar above the prompt as each plan step finishes, and a panel of the plan and every subagent's model, context, and cost.
Draws
Install
The marketplace name is garvit-mods.
/plugin install agent-monitor --marketplace GarvitNangru/claude-code-modsIt runs inside the Claude Code process. The side panel is only what the author's README names. Read the repo before you install, or run claude plugin validate on the plugin folder.
- Version
- 0.2.1
- License
- MIT
- Repo stars
- 24
- Last change
- 2026-10-11
What it can reach
- Draws in the session
- Reads usage or context figures
- Sees tool calls
The repository README says these mods were built and tested on Claude Code 2.1.292, on function hooks that can change between releases, and that the author has only tried them on Windows. Answer y to add the marketplace, then pick a scope. The user scope loads it in every session. It starts working right away, with no restart. When a task takes a few steps, the bar appears above the prompt on its own: the task, how many steps are done, what Claude is working on, and how long it has been going. If subagents start, the bar adds how many are running. When the last step is done the bar turns green and shows the total time. It clears once you send the next message. A quick question or a one-step request gets no bar. If Claude only runs subagents and has no plan, the bar counts those instead. The ✕ on the bar hides it, and Panel opens the full view. /agent-monitor opens the side panel and closes it if it is open. The panel is the plan as a checklist, then every subagent: its name, the model, the effort, how many model requests it has made, the context of its latest request and what share of the window that is, an approximate cost, and the wall-clock time. A finished agent, one you interrupted, and one that failed are marked differently. Clear finished drops the finished group from the panel. What the mod shows is kept for the current session only. In the desktop app the bar and the rows are graphics, with a small crab that walks while work is going on. Claude Code does not record how much of a task is done, so the mod gives Claude a tool named progress and adds two sentences to Claude's instructions explaining when to use it. Claude calls it once with the plan, then again as each step finishes. In the transcript the plan is one dim line. Later step updates add nothing there, because the bar already shows them. The README says you are never asked for permission, because the tool only draws the bar, and a rule you add that denies the tool wins. The author says the tool and the instructions add a few hundred tokens to each request, they are cached, and each step adds one short tool call. Claude decides what a step is, and sometimes skips the bar on a borderline task. The cost is an estimate from token counts and Anthropic's published API prices, with cache writes at the one-hour rate Claude Code uses. The README says it is not a bill, and that on a subscription you are not charged per token, so treat it as a rough weight. The price table is in the source. Update later with claude plugin update agent-monitor@garvit-mods, then /reload-plugins. Remove it with /plugin uninstall agent-monitor@garvit-mods.
The install name is agent-monitor. This page adds the author's name. The install line has no @suffix. The marketplace name is garvit-mods. The marketplace also holds readable-layout, ember, and firelight. This one installs on its own. The progress bar and the subagent panel are why this page files it under workflow. savvy-progress draws a progress bar above the prompt for /savvy-flow, and an agents panel with each subagent's model, task progress, context, cost, and time. The README says the idea comes from it, and that this one is a smaller take that works without any other plugin. todo-bar-hoobnn draws a progress bar of Claude's task list above the prompt: the task in progress, how long it has run, and what is next. This bar follows the progress tool the mod gives Claude. The two sentences and the progress tool are added to what Claude reads. The bar and the panel follow Claude's calls to that tool and the subagents Claude starts.
Demo

What it hooks into
session.startWhen a session startscommand.runWhen you run its slash command/agent-monitortool.describeA Claude Code eventtool.checkA Claude Code eventprompt.composeWhile you write a prompttool.callAround a tool call, where it can hold oneturn.startWhen Claude starts a turnagent.spawnWhen a subagent startsturn.stepAt each step of a turnturn.completeWhen a turn endsprompt.submitWhen you send a promptui.renderDraws into the sessionAbovePromptPane
Read from the on() calls in its hook modules. Open its source under Files to check what each one does.
Slash commands 1
/agent-monitorShow or hide the Agent Monitor panel: Claude's current plan and every subagent with its model, context, estimated cost and time
Files
The author's README and source, 4 files. They load when you open them.
Facts
- Author
- GarvitNangru
- Category
- Workflow
- Marketplace
- garvit-mods
- Folder
- mods/agent-monitor
- Version
- 0.2.1
- License
- MIT
- Repo stars
- 24
- Last change
- 2026-10-11
- Language
- TypeScript
- Files
- 7