MA

mikelewis1971/ai_workspace_mcp

开发工具

本页由 AI 翻译,技术细节请以原文为准。 查看原文

11 stars 4 forks 质量 74 趋势 22MIT

A swiss army knife for your local AI model. Your AI can now browse the web, run python scripts, save files, access data and so much more

概览

https://notebooklm.google.com/notebook/6bae6b7a-3c7f-40ba-9e09-43fbc82bd5a0 If you want to hear an audiopodcast about the security issues or just a deep dive into deployment, follow that link. Do you use LM Studio or Jan? Or even just Python and LLama.cpp? If so, you need this mcp server. You don't need any third party vendors. All the apps are selfcontained and private to your local system. All you need do is download LM Studio or Jan.ai, download a model (I highly suggest a Qwen 3.5 Claude version), copy and paste this into your mcp.json and reload your AI. Done. Well, you may have to debug for your particular system. This works on my system perfectly but it may not on yours. Never fear. If you don't know how to debug, simply download Anitgravity from Google and ask the model to read this repo, find your current copy of LM Studio or Jan and get the mcp.json aligned to your system. The Antigravity model from Google will fix any issues you may have.

README

以下为原文 README 节选,完整内容请查看原文。

https://notebooklm.google.com/notebook/6bae6b7a-3c7f-40ba-9e09-43fbc82bd5a0 If you want to hear an audiopodcast about the security issues or just a deep dive into deployment, follow that link.

Do you use LM Studio or Jan? Or even just Python and LLama.cpp? If so, you need this mcp server. You don’t need any third party vendors. All the apps are selfcontained and private to your local system. All you need do is download LM Studio or Jan.ai, download a model (I highly suggest a Qwen 3.5 Claude version), copy and paste this into your mcp.json and reload your AI. Done.

Well, you may have to debug for your particular system. This works on my system perfectly but it may not on yours. Never fear. If you don’t know how to debug, simply download Anitgravity from Google and ask the model to read this repo, find your current copy of LM Studio or Jan and get the mcp.json aligned to your system. The Antigravity model from Google will fix any issues you may have.

ai_workspace_mcp

1. The Full mcp.json for LM Studio

Create (or edit) this file in your LM Studio configuration folder.
LM Studio will open it automatically if you go to the Program tab → Install → Edit mcp.json.

{
  "mcpServers": {
    "ai-workspace": {
      "name": "AI Workspace MCP Server ",
      "description": "Self-upgrading, self-healing AI workspace with file ops, OCR, vision, mouse/keyboard control, evolving RAG, and full self-upgrade/restore capabilities. Built for LM Studio local models.",
      "command": "python",
      "args": [
        "F:/AI Sandbox/ai_workspace_mcp.py"
      ],
      "env": {},
      "workingDirectory": "F:/AI Sandbox",
      "enabled": true
    }
  }
}

Important notes for this config:

  • Change the paths only if your folder is somewhere else.
  • The command is python (it will use whatever python is in your PATH — the Windows Python that has the mcp package installed).
  • workingDirectory must match exactly where your ai_workspace_mcp.py lives.
  • After saving, toggle the server ON in the Program tab (or restart LM Studio).

2. Updated Installation Guide (replaces the previous MD)

# AI Workspace MCP Server 
**Self-Upgrading + Self-Healing AI Workspace for LM Studio & JAN.ai**

**Version:** Self-Upgrading (v2026.04)

---

## What Is This?

A complete, local-first MCP server that turns any LM Studio (or JAN.ai) model into a powerful, self-improving AI agent that can control your desktop, remember everything, upgrade its own code, and recover if anything ever breaks.

---

## Installation (LM Studio – Recommended)

### Step 1: Prerequisites
- Windows 10/11
- Python 3.10+ (added to PATH)
- WSL + Ubuntu 24.04 (optional but excellent for Conda)
- LM Studio (latest version – 0.3.17 or newer)

### Step 2: Create the Workspace
1. Create the folder: `F:\AI Sandbox`
2. Place your `ai_workspace_mcp.py` file inside it.

### Step 3: One-Time Setup
In **PowerShell as Administrator**:
```powershell
在 GitHub 查看完整 README

安装

This server does not publish a one-line install command.

Open the repository installation guide 新手?先看 mcpskill.org 教程 →

配置

{ "mcpServers": { "ai-workspace": { "name": "AI Workspace MCP Server ", "description": "Self-upgrading, self-healing AI workspace with file ops, OCR, vision, mouse/keyboard control, evolving RAG, and full self-upgrade/restore capabilities. Built for LM Studio local models.", "command": "python", "args": [ "F:/AI Sandbox/ai_workspace_mcp.py" ], "env": {}, "workingDirectory": "F:/AI Sandbox", "enabled": true } } }