local-computer-full-access-mcp
Provides full access to a local Windows computer via MCP, enabling arbitrary command execution, file operations, and process management through a secure stdio or HTTP tunnel.
README
Local Computer Full-Access MCP
A Windows-first MCP server that gives an MCP client Codex-like local capabilities through explicit tools:
- unrestricted PowerShell, Command Prompt, or Bash commands;
- read, create, edit, copy, move, search, and delete files/directories;
- list drives and inspect disk capacity;
- start, inspect, and terminate processes;
- local JSONL audit logging;
- stdio transport for local MCP hosts;
- Streamable HTTP plus an HTTPS tunnel for ChatGPT;
- DNS-rebinding protection with tunnel Host-header rewriting;
- one-click shutdown via
STOP_MCP.bat.
run_command makes this genuinely full access under the Windows account running the server. It is not a simulated filesystem.
[!CAUTION] This project can execute arbitrary commands, modify or delete files, start processes, and expose everything accessible to the Windows account running it. Use it only on a computer you control. Keep the server bound to
127.0.0.1, never publish the complete MCP endpoint URL, and stop the server when it is not in use. Read SECURITY.md before enabling full access.
ChatGPT connectivity and availability
ChatGPT cannot connect directly to 127.0.0.1 on your PC. This package keeps the MCP service on localhost and uses cloudflared to create an outbound HTTPS tunnel for testing.
Availability of custom MCP apps and write/modify actions depends on the ChatGPT plan and workspace configuration and may change over time. Check OpenAI's current developer mode and MCP apps documentation. The server also works with compatible local MCP hosts over stdio.
Install on Windows
-
Extract this folder to a permanent location, for example:
%USERPROFILE%\Documents\Local-Computer-Full-Access-MCP -
Double-click:
INSTALL_WINDOWS.bat
The installer creates a Python virtual environment, installs the official MCP Python SDK, installs cloudflared through winget when needed, creates .env, enables full access because that is the requested mode, generates a random secret endpoint, and runs tests.
Start for ChatGPT
- Double-click
START_CHATGPT.bat. - Keep the terminal window open.
- Wait until it prints
CHATGPT MCP ENDPOINT. - Copy the complete endpoint ending in
/mcp/<long-secret>. - In a supported ChatGPT workspace on the web, enable Developer mode and create a custom app/MCP connector.
- Paste the endpoint, choose No authentication for this temporary capability-URL setup, scan the tools, and create the app.
- Enable the app in the chat. Leave confirmations enabled for risky actions.
To stop everything, close the launcher window normally or run STOP_MCP.bat.
The TryCloudflare hostname changes whenever the script restarts, so update the custom app endpoint after each restart. TryCloudflare is for testing, not a permanent deployment. If cloudflared reports that a Quick Tunnel cannot start because you already have %USERPROFILE%\.cloudflared\config.yml or config.yaml, temporarily rename that file or use a separately configured stable Cloudflare Tunnel.
Start over stdio
Double-click START_STDIO.bat, or point a local MCP client at:
command: <folder>\.venv\Scripts\python.exe
args: -m pc_mcp.server
environment: PC_MCP_TRANSPORT=stdio
An example generic client config is in mcp-client-config.example.json.
Tools
Read-only tools:
pc_statuslist_drivesfile_infolist_directoryread_text_fileread_binary_file_base64search_fileslist_processes
Write/action tools:
write_text_filewrite_binary_file_base64replace_textmake_directorycopy_pathmove_pathdelete_pathrun_commandstart_processterminate_process
Example prompts after connection
- "Use my local-computer app to list my drives and show free space."
- "Inspect the selected project folder, run its tests, fix failures, and show me the changed files."
- "Search the selected project folder for references to a specified phrase without changing anything."
- "Create a backup of this folder before editing it. Ask before deleting anything."
Configuration
The installer creates .env. Important settings:
PC_MCP_FULL_ACCESS=1
PC_MCP_TRANSPORT=streamable-http
PC_MCP_HOST=127.0.0.1
PC_MCP_PORT=8765
PC_MCP_ENDPOINT_SECRET=<random-secret>
To restrict filesystem tools:
PC_MCP_FULL_ACCESS=0
PC_MCP_ALLOWED_ROOTS=%USERPROFILE%\Documents;%USERPROFILE%\Projects
See SECURITY.md before leaving this server running or converting the test tunnel into a permanent endpoint.
Test manually
Run RUN_TESTS.bat.
For MCP protocol inspection, activate the virtual environment and run:
.\.venv\Scripts\Activate.ps1
mcp dev server.py
Audit log
Tool actions and command outcomes that reach the core implementation are recorded locally at:
.data\audit.jsonl
File contents, binary payloads, and environment dictionaries are redacted and hashed in the audit record.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.