macbot-mcp CLAUDE.md

macbot-mcp CLAUDE.md is an instructions file for coding agents from anomalous3/macbot-mcp. It costs 999 tokens per session, scanned A, original, MIT.

A set of instructions for controlling macOS apps through their accessibility controls, AppleScript, and Quartz, including browser and other graphical interfaces.

In plain words
What is it for?
Use it to read interface trees, find buttons and fields, press controls, retrieve browser URLs, and use screenshots when visual details are necessary.
Why use it?
It helps an agent inspect and operate interface elements without relying on screenshots or mouse coordinates in many cases.

Instructions file

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add instructions/anomalous3/macbot-mcp/claude-md
Clone the repo
git clone --depth 1 https://github.com/anomalous3/macbot-mcp

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for macbot-mcp CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/anomalous3/macbot-mcp/claude-md.svg)](https://agentmods.dev/instructions/anomalous3/macbot-mcp/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/anomalous3/macbot-mcp/claude-md"><img src="https://agentmods.dev/badge/instructions/anomalous3/macbot-mcp/claude-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 999 This file is loaded in full into every session.
When invoked 999 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00999 $0.00999
Opus 5 $0.00500 $0.00500
Sonnet 5 $0.00200 $0.00200
Haiku 4.5 $0.00100 $0.00100

Measured 4d ago against content hash cb41d4234619, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

macbot-mcp CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

CLAUDE.md · 107 lines

How it starts

The opening of the file, as written. The whole thing — 107 lines — stays where its author put it; the contents beside it link to each section on GitHub.

macbot-mcp

macOS GUI automation via the Accessibility API, AppleScript, and Quartz.

When to use which tool

Prefer accessibility tools over screenshots. They're 5-15x cheaper in tokens and return structured, searchable data.

Goal Tool Tokens
Read UI state (windows, buttons, tabs, text) mac_ui_tree ~200-700
Find a specific element (button, link, field) mac_ui_find ~100-300
Get the browser URL mac_ui_url ~50
See what something looks like visually mac_screenshot + Read ~3000

Use screenshots only when you genuinely need visual information — layout, colors, images, charts, or anything the accessibility tree can't describe.

Common workflows

Press a button without a screenshot (or coordinates)

mac_ui_press(app="Firefox", name="Submit", role="AXButton")

One call: finds the element and presses it through the Accessibility API. No mouse movement, no focus change — works even if the window is occluded or the element is scrolled out of view. This beats find+click for buttons, links, tabs, and checkboxes; use index= to pick among multiple matches.

Click when press isn't supported (canvas, custom widgets)

1. mac_ui_find(app="Firefox", name="Submit")
   -> {"role": "AXButton", "title": "Submit", "x": 450, "y": 320, "width": 80, "height": 32}
2. mac_click(x=490, y=336, app="Firefox")

Pass app= to mac_click: a click lands on whatever window is at those coordinates, so an occluded target means the wrong window gets it.

Navigate browser tabs

1. mac_ui_tree(app="Firefox", max_depth=5, max_nodes=50)
   -> Find AXRadioButton elements in the tab group with titles and coordinates
2. mac_click at the tab's coordinates
3. mac_ui_url to confirm you're on the right page

Type into an app

1. mac_type_text(text="Hello world", app="TextEdit")
2. mac_key_press(key="return", app="TextEdit")

Always pass app= when typing. Do NOT split focus and typing into separate tool calls (mac_focus_app then mac_type_text) — seconds of model thinking elapse between calls, the terminal running the agent reclaims focus, and the text lands in the wrong window (usually your own input box). The app= param focuses and types atomically inside one call.

Read the full file on GitHub · 107 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 4d ago First seen · 107 lines · 999 tokens per session scan A cb41d4234619

Subscribe to this mod's changes

macbot-mcp CLAUDE.md is an instructions file published in the GitHub repository anomalous3/macbot-mcp (0 stars, last pushed 16d ago), licensed MIT. It adds 999 tokens to every session, about $0.0050 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.