Let Claude Code and Codex point at Mac screen elements with bigarrow
An open-source, MIT-licensed macOS command-line tool called bigarrow, with a skill for Claude Code and Codex, lets AI agents draw a click-through arrow and text sign over the screen. This lets an agent show a human where to click, for example on permission dialogs, instead of printing instructions into a terminal.
The problem
AI agents doing real work on a Mac often reach a step that only a human may or can do, such as clicking Allow in a permission prompt, passing 2FA, or pasting a code. The agent usually just prints a request into a terminal the person is not watching. Nothing guides the person to the right button or control on screen.
How they did it
- Install the tool with Homebrew: brew install franzenzenhofer/tap/bigarrow (or build from source with Swift, Xcode 16+, macOS 14+).
- Run bigarrow install-skill to add the skill for Claude Code and Codex.
- The agent picks a target by label (--element with --app), by coordinate (--at), or by window (--window).
- The agent writes a full sentence on the sign, optionally with a {{value}} copy button or spoken aloud via --say.
- The arrow removes itself after a time limit, on stop, when the agent process exits, or when the human clicks the sign or its close button.
From the official docs
bigarrow point --element "Allow" --app "System Settings" --text "Franz, click Allow: Ghostty may control your Mac"Results
- Drawing needs no macOS permission; only some target-finding modes need Accessibility or Screen Recording, granted to the terminal app that launched bigarrow.
- The README reports 104 automated tests and 18 behaviour checks on a clean CI runner, covering click-through, focus never moving, multiple displays and Spaces.
- The README reports the skill's description costs 182 tokens, with the instructions at 1,528 tokens loaded only when the agent decides to point.
- As reported by the project README, the repository had 611 stars and 11 forks at capture time.
As reported by the source (GitHub repository README (franzenzenhofer/big-arrow-on-the-screen)); AgentGid did not measure these figures.
This suits solo developers or small teams already running Claude Code or Codex on a Mac who often leave agents unattended. Setup is one Homebrew command, though target-finding modes need Accessibility or Screen Recording granted to your terminal. The test counts and token costs come from the project's README, so check them yourself. Pi (Free, OSS) is a cheaper agent option, but the case doesn't say whether the skill works there.
The agents used here
Anthropic's agentic coding tool that reads a codebase, edits files and runs commands from the terminal, with VS Code and JetBrains integrations plus …
- Price
- $20/mo
- Gid Score
- 81 #2 in Terminal
OpenAI's coding agent, offered as an open-source terminal CLI, an IDE extension, a desktop app and a cloud agent that works on repositories. Users si…
- Price
- Free + $8/mo
- Gid Score
- 84 #1 in Terminal
Compare: Claude Code vs OpenAI Codex
Similar use cases
Route between coding-agent models with the open-source Weave Router 2.0
The Weave team presented Weave Router 2.0 in a Show HN post. It is an open-source routing model that sits in front of coding agents and switches between LLMs per turn, and the aut…
As reported by the Weave author on Hacker News: on Terminal Bench 4.0 and SWE Atlas, the router had equivalent pass rates to GPT-6 Astra.
Organize agent skills as installable company departments with headcount
An open-source GitHub repository called headcount arranges agent skills like a company, with a chief executive over departments such as security, finance, product and marketing. E…
As reported by the repository README: a chief executive over 16 departments, 172 skills in total.
Run Claude Code and Codex agents in parallel on Mac with Offrun
Offrun is a free Mac app that runs coding agent CLIs such as Claude Code and Codex side by side, each in its own git worktree, using the user's own logins. The page is the product…
As reported by the Offrun site: each agent in a project works in its own git worktree, so two agents on the same repo do not touch the same files.
Guides
Claude Code Deploys: 6 Open-Source Repos for a Containment System, Checked
A viral post says GitHub Actions, Argo CD, Argo Rollouts, Prometheus, Sentry and OpenTelemetry turn Claude Code deploys into a containment system. We checked each: role, license, what needs Kubernetes, and what Claude Code's own permissions add.
How to Cut Token Usage in Claude Code and Cursor (2026)
Practical ways to spend fewer tokens with Claude Code and Cursor: measure first, keep context small, match the model to the task, filter what the agent reads, and the open-source tools that help.
Sonnet Builds, Opus Reviews, Haiku Reads: Does the Claude Code Advisor Save Money?
A viral post says running Sonnet, Haiku and Opus together in Claude Code gives better code for fewer tokens. We checked it against Anthropic's docs and prices: when the advisor pays off, when it costs more, and why a ready project kit matters more than the model mix.
JEV for Claude Code: A Second Opinion on Every Command, Checked
A viral post says jev-axi screens every command your Claude Code agent runs for a fraction of a cent. We read the repo, TypeSafe's docs and privacy policy: what runs locally, what is sent, how it fails, and how it compares with auto mode.