Sourcery MCP
Let your coding agent request Sourcery reviews and fix the comments and security findings itself
agent-desktop
Computer use for any coding agent on macOS, Windows and Linux
About
agent-browser, for your whole computer
agent-desktop is built in Rust and ships 56 operational commands: snapshots, clicks, typing, keyboard and mouse input, app and window management, notifications, clipboard and screenshots. Refs belong to the snapshot that produced them, so a stale UI returns STALE_REF instead of a wrong click, and every command prints one JSON envelope with an error code and a retry hint. Ref actions are headless by default: they go through the accessibility API without moving the pointer or taking focus, so it runs in the background while you keep working. The launch post says several agents can use the computer in parallel, each with its own cursor. The binary carries its own agent docs (`agent-desktop skills get desktop`), and a skill is available for Claude Code, Codex, Cursor and other harnesses through skills.sh. macOS needs Accessibility permission, Linux needs AT-SPI. Built by Vercel Labs, lead developer Lahfir ([@mdlahfir](https://x.com/mdlahfir)), announced by Chris Tate ([@ctatedev](https://x.com/ctatedev/status/2116323148706771302)). Repo: https://github.com/vercel-labs/agent-desktop.
Features
- Accessibility tree snapshots with stable refs instead of pixel guessing
- macOS, Windows and Linux (AT-SPI), x64 and ARM64
- Headless by default: acts without moving the pointer or taking focus
- 56 commands for apps, windows, clipboard, notifications and screenshots
- JSON output with error codes and safe-retry hints
- Agent skill for Claude Code, Codex and Cursor, plus a C library for in-process use
Install
Related skills
All skillsRelated articles
All articlesGrok Bot vs Claude Code and Codex: What Each One Is For
How Grok Bot, a persistent personal agent, differs from coding agents like Claude Code and Codex, and when it makes sense to use each one or both together.
Browser Automation for AI Agents (2026)
I compare five ways to give an AI agent a browser, from local Chrome over MCP to hosted cloud browsers, with real install commands and pricing.
7 Tools to Cut Your Coding Agent's Token Bill (2026)
Seven tools that cut Claude Code, Codex and Cursor token usage by compressing shell output, searching code and fetching current docs, with honest caveats.
