Markdown version of https://extraheadroom.com/headroom-vs-ponytail

# Headroom vs Ponytail vs RTK: what's the difference?

Headroom, Ponytail, and RTK each cut the token cost of Claude Code and ChatGPT Codex on a different part of the bill, and they run together in the same app, so there is no need to pick one. This page covers what each does and when it matters.

## Quick answer

They are not competitors. **Headroom** trims what goes _into_ the model (bloated tool output, logs, and boilerplate) before it reaches Claude Code or ChatGPT Codex. **Ponytail** trims what comes _out_: it nudges the agent to write the least code that solves the problem, cutting output tokens. **RTK** trims noisy terminal command output, and **MarkItDown** converts heavy documents to clean Markdown. Ponytail, RTK, and MarkItDown are one-click add-ons inside the Headroom app, so most people run all of them at once.

## Side by side

| Tool | What it trims | Side of the bill | How you get it |
| --- | --- | --- | --- |
| Headroom | Bloated tool output, repeated context, logs, and boilerplate before they reach the model. | Input | The core app (or the [open-source CLI](/headroom-desktop-vs-cli)). |
| Ponytail | The code the agent writes back: less code, fewer output tokens, less to maintain. | Output | One-click add-on / skill. |
| RTK | Noisy terminal command output the agent would otherwise read in full. | Input | One-click add-on. |
| MarkItDown | Heavy documents (PDF, Office) converted to clean, compact Markdown. | Input | One-click add-on. |

## Headroom vs Ponytail

They solve different halves of the same problem. [Headroom](/) works on the input side: it compresses the repetitive tool output and boilerplate piling up in the context window, locally and forwards a smaller request to the model. [Ponytail](/ponytail-claude-code) works on the output side: it gives the agent a "lazy senior developer" rule set so it ships the least code that works, which costs fewer output tokens to generate and less to re-read and edit later.

Because one trims what goes in and the other what comes out, their savings add up rather than overlap.

## Where RTK and MarkItDown fit

RTK and MarkItDown are narrower input-side tools that handle two specific sources of bloat. RTK trims the noisy output of terminal commands so the agent reads a compact summary instead of hundreds of lines. MarkItDown converts heavy documents (PDFs, Office files) into clean Markdown before they hit the context window. Both are one-click add-ons in the Headroom app, alongside Ponytail.

## Which should you use?

For most developers the answer is all of them, because they stack rather than overlap. Headroom is the core input-side compressor; Ponytail, RTK, and MarkItDown each remove a different kind of waste on top. Together they target up to ~50% lower token usage: up to 2x more work on the plan you already pay for. The simplest path is to [download the Headroom app](/) and turn on the add-ons you want from one screen. See the [install guide](/install-headroom-claude-code) for the app-vs-CLI details, or [what's new](/docs/changelog) for the current add-on lineup.