agent-top

htop for local coding agents.

You have three Claude Code sessions, a Codex thread in VS Code, and a Gemini CLI you forgot about. Which one is burning tokens right now? Which one is waiting on you? Which MCP server is still alive after the agent that started it died? agent-top answers that in one terminal view, the way htop answers it for processes and btop answers it for the whole machine.

Recorded from a synthetic snapshot (docs/demo-snapshot.json, replayed with --replay) rather than a live machine, because a recording of real sessions would publish real project names, working directories and session ids. Regenerate with vhs docs/demo.tape.

Quick start

brew install kannandreams/tap/agent-top   # or: cargo binstall agent-top
agent-top

That is the whole setup. There is nothing to configure and nothing to enable in your agents: agent-top reads the transcripts the harnesses already write and the process table the OS already keeps. Start it in any terminal while your agents run.

What to look at first:

• STATE tells you who is working and who is waiting for you.
• COST is what each session has spent so far, at list price.
• Red rows in the detail pane are MCP servers whose agent has gone. They are the leak this tool exists to catch.

Keys: j/k move, Tab switches the detail pane between the process tree and the tool trace, s sorts, x hides stopped sessions, ? shows the rest, q quits.

Other ways to run it:

agent-top --once             # print the table once and exit
agent-top --json             # one snapshot as JSON, for scripts and bug reports
agent-top trace --session 662cda1f -o trace.json   # one session as a trace file for Perfetto
agent-top --prices           # the price table in use, and where each row came from

What the table shows

A Claude Code session's subagents are folded into its row, the way Claude Code's own cost display counts them. Web searches the model ran are counted and, for Claude Code, priced.

The detail pane

Press t to open it and Tab to switch between two views.

Process tree. Every process under the agent, labelled agent, subagent, mcp, shell or tool, with the token breakdown beside it. Orphaned MCP processes, servers with no live agent above them, are listed in red.

Tool trace. A waterfall of the session's recent activity on a shared time axis:

 tool trace   5 of 71 calls · window 1m00s
   in tools 58%  model 31%  turn 3m20s…  slowest Bash 20.0s  1 in flight  1 failed
 model            4.2s  ▉▉▉▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏
 Bash             2.5s  ▏▏▏▉▉▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏
 ↳Grep           12.0s  ▏▏▏▏▉▉▉▉▉▉▉▉▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏
 Edit            300ms! ▏▏▏▏▏▏▏▏▏▏▏▏▉▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏
 Bash            20.0s… ▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▏▉▉▉▉▉▉▉▉▉▉▉▉▸

Each row is one tool call or one stretch of the model thinking (model). Width is its share of the window; colour is how long it took, green under a second through amber to red near a minute. ↳ marks a subagent's call, … one still running, ! one the harness reported as failed. in tools and model say how much of the window went to each; what is left is usually waiting on you. turn is how long the current human turn has been going.

None of this needs telemetry switched on. The spans are reconstructed from the transcript by pairing each tool call with its result and each prompt with its reply, reading only names, ids and timestamps.

Exporting a trace

agent-top trace --session 662cda1f -o trace.json

writes the whole session, every tool call, inference and turn, as a Chrome trace event file that ui.perfetto.dev and chrome://tracing open directly. Turns, tool calls and model time sit on separate tracks, main agent and subagents apart, so a turn shows as a bar with its calls beneath it. --session takes a session id, a unique prefix of one, or a path to a transcript file, and works on sessions that ended long ago. It writes a file and never contacts a collector.

Prices

Prices are data, not code. The table shipped in the binary lives in crates/agent-top-core/prices.toml, and a file of your own is merged over it at startup:

# ~/.config/agent-top/prices.toml   (USD per million tokens)

[[model]]
prefix = "gpt-5-codex"
input = 1.25
output = 10.0
cache_read = 0.125

An entry whose prefix matches a built-in one replaces it, so a price that has gone stale can be corrected without waiting for a release. A new prefix is added, which is how the models this project does not ship prices for get costed at all. Cache writes default to Anthropic's multipliers of the input price (1.25x for the 5 minute TTL, 2x for the hour) and can be set explicitly with cache_write_5m and cache_write_1h.

The longest matching prefix wins, so claude-fable-5-1 beats claude-fable-5, and a date-suffixed id like claude-sonnet-4-6-20251114 resolves to its base model. agent-top --prices prints the effective table with the source of every row, which is the quickest way to find out why something is showing n/a. A price file that cannot be parsed is reported on stderr and ignored; the built-in prices still apply.

A model with no entry anywhere is never guessed at. Its tokens are counted and reported as unpriced, and any total containing them is shown as a floor.

Web searches are billed per search, on top of tokens, and the rate is in the same file under [server_tools]. Codex web searches are counted but not priced, because OpenAI's rate is not in the table.

Supported harnesses

Install

Every route ends at the same single binary: no Python, no Node, no daemon. To upgrade, brew upgrade agent-top or re-run the cargo install command. Without Homebrew, agent-top --completions zsh (or bash, fish) prints a completion script to source from your shell's startup file.

All the flags

agent-top                        # interactive, refreshes every second
agent-top --interval-ms 500      # faster refresh
agent-top --stopped-window-min 120   # keep stopped sessions visible for two hours
agent-top --replay snap.json     # render a saved --json snapshot, keys and all, reading nothing local
agent-top trace --session <id|prefix|path> [--format chrome] [-o FILE]

--replay is for bug reports: attach a --json snapshot and the reader can inspect it exactly as you saw it.

Why this exists

The failure that motivated the tool is not hypothetical. Codex has a run of reports about leaked MCP process trees, three of them still open:

Nothing about this is specific to Codex. Every harness that spawns helper processes has the same shape of bug available to it, which is why agent-top looks for the symptom rather than for one vendor's bug.

Where the numbers come from

The whole point of this tool is that its numbers are right, so it is explicit about which ones are exact and which are inferred.

• Tokens are counted, never estimated. They come from the usage records the harness writes itself, deduplicated per API message so a response split across several transcript lines is counted once.
• Costs come from a table you can read and change. agent-top --prices shows it. A model with no price is reported as unpriced rather than guessed at, which is why a total containing one is shown as a floor (≥, +) instead of a number that looks more precise than it is.
• Attribution says how confident it is. Claude Code publishes a per-pid registry, so a session is matched to its process exactly. Codex has no equivalent, so the match is made on working directory and start time, and the detail pane labels that row a heuristic rather than presenting it as fact.
• Only metadata is read. Token counts, model ids, tool names, timestamps. Never a prompt, a tool input, or a tool result.
• Nothing is written, signalled, or sent anywhere. agent-top never kills or writes to an agent and makes no network calls. Killing an orphaned MCP server is your decision, with your own kill.

docs/architecture.md has the mechanism underneath: the process walk, the incremental transcript tail, and how a snapshot is assembled on each tick.

Roadmap

See docs/roadmap.md. Next: OTLP export, a Gemini CLI adapter, and per-MCP-server rows. docs/releasing.md is the release runbook.

Development

cargo test
cargo run -- --once
cargo clippy --all-targets

Two crates, split by dependency rather than by size: crates/agent-top-core is discovery, transcript parsing, pricing and the process model, with no terminal dependency, so all of it is testable without a TTY and it is exactly what --json prints; crates/agent-top is the ratatui front end and the CLI. Both are published, because a crate on crates.io cannot depend on an unpublished one: agent-top-core exists on the registry so that agent-top can. The internal engineering handbook (PRD, RFCs, ADRs, decisions) lives in the sibling agent-top-internal-docs repository.

License

MIT