agent-top

What is agent-top?

agent-top is a terminal monitor for the AI coding agents running on your own machine, in the spirit of htop for processes. It combines the transcripts that each harness already writes with the operating system's process table to build one table of sessions from Claude Code, Codex, Gemini CLI and OpenCode. For each row you can see whether the agent is working or waiting for you, how many tokens it has used, what it has cost at list price, and which helper processes it left behind. Red rows flag MCP servers whose agent has gone, a leak the tool checks for on every refresh. Codex rows also show the rate-limit windows the plan reports. Built in Rust, it comes as a single static binary for macOS and Linux and needs no configuration or daemon. It only reads, never writes to or kills an agent, and keeps your data on the machine. The site lists no pricing.

Use cases

• Monitor several coding agents at once
• Track token and dollar spend per session
• Find leaked MCP server processes
• Check Codex rate-limit usage
• Report cost by harness, model or project
• Capture a session snapshot for a bug report

Who it's for

• Developers using coding agents
• Engineering teams
• Terminal users

Why use agent-top

• See all your running coding agents in one place instead of switching between terminal windows.
• Know which session is idle and waiting for input, so you spend less time hunting for it.
• Watch spend per session as it accumulates, using token counts taken from the harness's own records rather than estimates.
• Catch leftover MCP servers that keep running after their agent exits.
• Your prompts, costs and process list are never sent anywhere, and the tool never signals or kills an agent.

agent-top features

• One table of all local agent sessions, refreshed every second
• STATE column showing who is working and who is waiting
• Token counts read from the harness's own usage records
• Cost per session at published list prices
• Editable price table through a personal prices.toml file
• Detection of orphaned MCP server processes
• Codex rate-limit windows with 75 percent and 90 percent colour warnings
• Cost report by harness, model, project or day
• JSON snapshots and replay of the whole screen
• Single static Rust binary with no runtime

agent-top integrations

• Claude Code
• Codex
• Gemini CLI
• OpenCode
• Kodelet

How to get started with agent-top

1. Install it with brew install kannandreams/tap/agent-top, or with cargo binstall agent-top.
2. Run agent-top in a terminal while your coding agents are working in other windows.
3. It reads the transcripts the harnesses write and the OS process table, and refreshes the table once a second.
4. Select a row to open the detail pane with state, tokens, cost, context by source and any leaked MCP servers.
5. Use agent-top --once for a one-shot listing, or export a snapshot to share for a bug report.

agent-top limitations

• Runs only on macOS and Linux, on the machine where the agents run
• Costs are list-price estimates from usage records, not an invoice
• Rate limits are read only from Codex; other harnesses show none
• A model with no price entry is shown as unpriced, so totals may be a floor

agent-top pricing

We couldn't confirm pricing on the official site. Check agenttop.dev for current prices.

Prices can change and may vary by country or exclude tax. Check the official pricing page.

agent-top alternatives

The main agent-top alternatives listed on Debuly are Claude Code, OpenCode, WhereMyTokens, AgentGrid, Poirot and Token Optimizer. The table compares pricing and platforms; order is A–Z.

1. 

   AgentGrid

   Desktop app that lays out Claude Code, Codex and other coding agents on one canvas, with usage tracking and account auto-switching.

   CodingProductivityAutomation and agents
   FreemiumVisit AgentGrid (official site)
2. 

   Claude Code

   Anthropic's coding agent that fixes bugs, writes tests and runs migrations from your terminal, IDE, desktop app, web or Slack.

   CodingAutomation and agentsDeveloper platforms
   PaidVisit Claude Code (official site)
3. 

   OpenCode

   Open-source AI coding agent for terminal, desktop and IDE, with optional low-cost Go model subscriptions and a team Console

   CodingDeveloper platformsAutomation and agents
   FreemiumVisit OpenCode (official site)
4. 

   Poirot

   Free, open-source macOS app that shows what Claude Code did in each session, with usage analytics and cost views.

   CodingDeveloper platformsProductivity
   Open sourceVisit Poirot (official site)
5. 

   Token Optimizer

   Source-available plugin that audits wasted context in Claude Code and other coding agents, compresses output and saves session checkpoints.

   CodingDeveloper platforms
   FreeVisit Token Optimizer (official site)
6. WM

   WhereMyTokens

   Local-first Windows tray app that tracks Claude Code, Codex and Antigravity token use, quota windows and cost estimates.

   CodingProductivityDeveloper platforms
   Open sourceVisit WhereMyTokens (official site)

agent-top FAQ

Sources & verification

Facts on this page come from the official sources below, each with the date we read it. Written by Debuly; the vendor did not review this page.

Image credits

• Logo: from agenttop.dev · fetched 23 Oct 2026 · Vendor trademark, used only to identify the product; removed on request
• Cover: none collected yet, so a generated tile is shown.

Spot an error? Report an issue with this listing. We also remove a logo or image on the owner's request.