"Google Pub/Sub" matching MCP servers:

Aggregates multiple local stdio MCP services into a single HTTP endpoint and launcher-managed local processes, with a mobile-first web console for starting, stopping, monitoring logs, health checks, and usage.

A compact facade MCP server for macOS desktop automation, lazily proxying to full-featured computer-use tools to save context tokens while enabling screenshots, window control, app launching, and AppleScript execution.

Enables AI assistants to autonomously control Windows 11 and 10 desktops via sub-10ms screen capture, native UI Automation element inspection, and zero-lag keyboard and mouse input. Combines a visual plane with a semantic plane and stall detection so agents can operate real applications reliably without vision-only guessing.

Enables AI assistants to scan local disks for developer caches, large unused videos, and old logs, present candidates for approval, and move only approved items to the operating system trash so they can be restored while protecting system folders.

Enables MCP clients to run Google Antigravity CLI (agy) as durable, detached asynchronous jobs with a permission gate and broker-computed verdicts that do not trust agy's own exit code.

Lets Claude see an Android phone's screen through ADB screenshots and control it by typing over ADB and by speaking button commands aloud through the PC speakers for Google Voice Access to tap, plus gestures, app launching, and optional voice input.

An unofficial experimental local MCP server that lets AI clients work with an already logged-in Windows Weixin desktop client, providing bounded conversation and message reads, group and message search, polling of up to 16 selected chats, guarded text and attachment submission, and image/file viewing handoffs. It also exposes hybrid Computer Use prompts for native mention handoffs, while remeasuring the window, DPI, and target controls on each action and failing closed when evidence is insufficient.

A cross-platform MCP server for capturing screenshots of monitors and windows, as well as listing and closing windows.

Enables AI agents to see, find, and click real UI elements on Windows with sub-pixel accuracy across mixed-DPI multi-monitor setups, even in apps that expose no accessibility tree, by routing requests across Win32, UIA, OCR, and pixel/template backends. Also provides window control, clipboard and recipe-driven automation, quiet clicks that avoid moving the cursor or changing focus, and built-in setup/diagnostic tooling for Claude Code, Claude Desktop, and Codex.

Enables loading plain-Python tool sets for Google Workspace (Gmail, Calendar, Chat, Docs, Sheets, Slides, Drive), Mac control, and clock/time-zone lookups, then exposing them as MCP tools with confirmation gates on sends and deletes. Tool sets can be served in-process to the Claude Agent SDK or over stdio to any MCP client, either individually, in groups, or all at once.

Set up Android TV, Google TV and Fire TV boxes over ADB on the local network: discover and pair boxes, set any launcher as the HOME app, install apps, take screenshots and run checks. No raw shell tool; the tools that change the box ask before they run.

Enables AI agents to drive an Android TV or Google TV box end to end through named tools — discover, pair, connect, check, install, provision, link, test, screenshot, report and a gated shell — with usage instructions embedded in the tool descriptions so the agent never improvises adb. Every command passes the same safety gate that refuses destructive actions.