Give the agent "this" and he gets a real PC.
This tool is for personal automation and testing on your own machine. Do not use it for unauthorized access.
Give the agent "this" — and it gets a real PC. Screen, mouse, keyboard, your browser, files, terminal, SSH — via the model's native tools.
RU README: Russian version · Skills: SKILL.md · IDE matrix: docs/IDES.md
- What it is - Why AiPC - Quick start - The aipc menu - 63 tools - Token efficiency - MCP connection - Self-update - Safety - Troubleshooting - Project structure - Build from source - License
AiPC is a local service + console utility that gives any AI agent full access to your computer over the open MCP (Model Context Protocol) standard.
The agent stops being "text in a chat" and starts working like a human at the PC: it looks at the monitor, clicks, types, opens apps, googles in your browser, runs commands, uses SSH — and verifies every step with a new screenshot.
If a model says "I have no computer access", the built-in system prompt (aipc/server.py:SYSTEMPROMPT, also served as the aipcinstructions MCP prompt) forbids that answer and requires acting through AiPC tools.
Usual agent limits With AiPC ------------------------------- Blind: no screen screensee — screenshot as a native image block, cursor marked Clicks by guessing uisnapshot / uifind — ready-made x/y coordinates (0–1000) Typing into the void focustype — verified focus + atomic typing; refuses to type blind Sleeps and hopes waitforwindow / waitforuielement / waitforchange "Did it work?" unknown screenshotdiff, assertui, logstail, audit.log No machine access Files, terminal, processes, SSH/S
From the project README.
Add the radar badge to your README — it shows your project was picked up by MCP Radar and links to this page:
[](https://mcp.liqiwa.com/s/SysikNagibator--AiPC.html)
Open-source, LLM-agnostic capability runtime for AI agents — 97 skills as typed, permission-enforced MCP tools. Final certification: 23/23 gates PASS.
JUSICK/Argos-mcp-guardrailA lightweight, sub-millisecond security shim that prevents AI agents from accessing sensitive files (.env, SSH keys) and executing destructive commands via MCP.
0xwilliamortiz/openclaude-improvedruns anywhere. uses anything
junxchan54-maker/mcd-cn-assistant🍟 麦当劳中国官方 MCP 智能助手技能：把 35 个原子工具编排成 4 条业务流 —— 活动推送 · 联名关注 · 最优惠点餐 · 多人群点聚合账单。算价用 calculate-price 真算，下单必二次确认。｜McDonald's China MCP agent skill: campaigns, collab watch, best-value ordering, group bill s
metahub-ai/assayAn open, reproducible trust layer for AI artifacts — skills, MCP servers, agents, and plugins. Assay inspects code, runs sandboxed behavioral audits, and generates verifiable reports. Help us build th
modulastack/modula-relayLocal MCP relay for cross-model agent teams. Your Claude, Codex, and other agents talk to each other over Unix sockets. No broker, no network, no accounts.
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.