automation

typesafe-computer-use

Computer use for about $0.0002 a step: OCR the screen, classify the next action with TypeSafe, click. macOS.
1,261 stars+26★ this week120 forksPythonUpdated 9/29/2026100% free · open source
What it does

Runs a low‑cost, AI‑guided loop that reads the screen (OCR), decides the next UI action with the TypeSafe model, and clicks – letting you automate macOS GUI tasks for pennies per step.

When to use it
  • •You need to click through a web‑app that has no API (e.g., upload a file on a legacy portal).
  • •You want to scrape data from a desktop app that only shows information on screen.
  • •You have a repetitive UI workflow (e.g., generate a report in Excel, copy‑paste into another app) and want to automate it without writing AppleScript.
Ready-to-paste prompt
python -m typesafe_computer_use run - <<'EOF'
goal: "Open Safari, navigate to https://app.company.com/reports, click the ‘Export CSV’ button, and save the file to ~/Downloads/report.csv"
EOF
Heads up: The script will fail unless the Python interpreter has macOS Accessibility permission – you must add it in System Settings before the first run.
Saves to your device
Use with Claude
New

Skip the builder — one click puts this in Claude, Cursor, Antigravity and more.

🔑 Needs an API key

Works after you add credentials — the setup agent will ask you for them.

Try it instantly — no install
Claude Code
mkdir -p ~/.claude/skills/typesafe-computer-use && curl -fsSL https://workflowstacks.com/api/skills/typesafe-computer-use/claude-skill -o ~/.claude/skills/typesafe-computer-use/SKILL.md
Open in another AI app

Opens the app with this repo with the prompt ready to go — no copy-paste needed.

Connect the whole catalog (MCP)
claude mcp add --transport http workflowstacks https://workflowstacks.com/api/mcp

Adds a WorkflowStacks connector to Claude Code: search and load any skill here by chatting.

How typesafe-computer-use works
Codeflow
Free to inspect

Typesafe-computer-use is a medium Python project (~12k lines across 45 code files, plus 33 test files). Setup is light: installs like a normal app — add your API keys. Reading the code is optional. Last commit a month ago, MIT license, has a test suite.

Size
Medium codebase
~12k lines · 45 code files · ~2 h to skim
Setup
One-command install
Installs like a normal app — add your API keys. Reading the code is optional.
Runs on
Python · Docker
Needs API keys (.env)
Python 92%Shell 6%HCL 2%
Where to start reading
  1. 1
    README.md
    Start here — what it does and how to install it
  2. 2
    AGENTS.md
    The instructions the AI actually follows
  3. 3
    typesafe_computer_use/cli.py
    Where the program starts running
  4. 4
    pyproject.toml
    Dependencies and the commands it exposes
  5. 5
    .env.example
    The API keys and settings you must provide
What's in each folder
typesafe_computer_use/Core code — the actual logic38 files
docs/Documentation10 files
benchmarks/Evaluations & benchmarks8 files
infra/Deployment / infrastructure6 files
osworld_overlay/Folder4 files
sandbox/Folder4 files
scripts/Helper scripts3 files
bench/Folder1 files
READMEHas testsDocumentedCI checksDocker readyMIT licenseUpdated this month
Quick Actions
Details
Creator
awlevin
Language
Python
Category
automation
Published
9/16/2026

Are you the creator of this tool? and earn 85% of every sale.

Show it off in your README:

Featured on WorkflowStacks
[![Featured on WorkflowStacks](https://workflowstacks.com/api/badge/typesafe-computer-use.svg)](https://workflowstacks.com/skills/typesafe-computer-use?utm_source=github&utm_medium=badge)
🔥 Hot this week, in your inbox
Skills like this, every Monday.

The five fastest-growing open-source AI skills, ranked by GitHub star growth. One email a week, unsubscribe anytime.