Easy Local MCP
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Easy Local MCPunlock and show me the files in my Documents folder"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
What is Easy Local MCP?
Easy Local MCP connects ChatGPT to capabilities on your own computer through the Model Context Protocol (MCP).
flowchart TD
A[ChatGPT] -->|MCP| B[Cloudflare Relay / Worker]
B -->|WebSocket| C[Easy Local MCP Agent]
C --> D[Files]
C --> E[Shell]
C --> F[Processes]
C --> G[Skills]
C --> H[External MCP]The Agent makes the outbound connection, so you do not need a public IP address or an inbound port.
Easy Local MCP is based on daodao97/localmcp and is maintained as an independent project.
Related MCP server: GPT Agent
Highlights
Windows desktop app with Control Center and system tray
Files: browse, search, read, create, edit, move, and delete
Shell commands and persistent processes
Multiple workspaces
Skills
External MCP servers through a stable gateway
Safe-by-default permission model
Local LOCK / UNLOCK gate for privileged operations
Authenticated local IPC
Audit logging and credential rotation
Public Relay for quick setup, with self-hosted options for long-term or sensitive use
Bundled Node runtime in the Windows installer
Install
Windows
Download the latest Windows x64 installer from GitHub Releases.
The desktop installer includes the required runtime. The target PC does not need Node.js or Rust.
Build from source
Requirements:
Node.js 22+
Rust stable for the desktop app
Visual Studio 2022 / MSVC and Windows SDK for the Windows installer
git clone https://github.com/Ryanma-YX/easy-local-mcp.git
cd easy-local-mcp
npm ci
npm run desktop:bundleQuick start
Start Easy Local MCP.
Choose the public Relay for a quick start, or enter the URL of a Relay you control.
Click Save & Start Agent.
In Control Center, use Reveal / Copy MCP URL.
Enable ChatGPT Developer Mode: Open Developer Mode settings.
Create a connector: Open Create Connector.
Paste the complete MCP URL and set Authentication to None.
Start a new chat and use the connector.
ChatGPT may hide the Create Connector entry in the normal UI. The direct link above opens the connector creation page.
The complete MCP URL contains an access credential. Do not post it in issues, screenshots, logs, chats, or Git repositories.
Security model
The Agent starts LOCKED.
Configured privileged tools remain visible in the MCP tool list while locked so ChatGPT can discover and refresh connector capabilities correctly. Visibility does not grant execution permission.
While LOCKED:
read-only capabilities remain available when enabled
configured privileged capabilities remain discoverable
file writes, deletes, shell commands, processes, and external MCP execution are denied
Privileged operations become callable only after a local Unlock and are blocked again when the Unlock expires or the Agent is locked.
For the full security model, see SECURITY.md and the Security Model wiki page.
Common commands
easy-local-mcp ui
easy-local-mcp status
easy-local-mcp url
easy-local-mcp unlock
easy-local-mcp lock
easy-local-mcp reload
easy-local-mcp rotate
easy-local-mcp join <zone-code>
easy-local-mcp stopThe legacy localmcp command remains available for compatibility.
Self-host the Relay
The prefilled public Relay is currently provided as an upstream/community shared service and is convenient for quick evaluation. Because it is operated outside this fork, this project cannot guarantee its availability, capacity, or long-term continuity.
For regular use, company source code, ERP/MES data, internal files, credentials, or production environments, a Relay under your own control is recommended. You can host it on a machine you manage (with a reachable HTTPS endpoint), on your own VPS, or on your own Cloudflare Worker.
After deployment, copy the Relay or Worker URL and configure it in Easy Local MCP.
See Self-hosted Relay for deployment and registration-protection details.
Multi-Device Zones
A Relay can group several devices into a Zone. The /admin dashboard is protected by a dedicated RELAY_ADMIN_TOKEN_HASH; without that configuration it fails closed and exposes no management actions. After Relay Admin login, you can create/import Zones, generate one-time join codes, view device status, rotate the shared connector, rename devices, or revoke them. On a stopped device, stage the join with localmcp join <zone-code>, then start Easy Local MCP normally.
A Zone also exposes one shared MCP connector URL. ChatGPT can connect to that URL once, call list_devices, and route the normal LocalMCP tools by adding a required device argument. Per-device MCP URLs continue to work for compatibility, and the target Agent still enforces its own local LOCK and feature permissions. Existing Zones created by an earlier Zone build can enable the shared connector by choosing Rotate Connector in /admin. See Multi-Device Zones.
Development
npm ci
npm run check
npm test
npm run buildDesktop-related commands:
npm run tray:check
npm run tray:build
npm run desktop:bundleMore documentation is available in the project Wiki.
Project origin
Easy Local MCP is derived from daodao97/localmcp. Thanks to the original project for the Local MCP and Relay foundation.
This fork is maintained independently and is not intended to merge back upstream.
License
MIT. See LICENSE.
This server cannot be deployed
Maintenance
Related MCP Connectors
MCP connector that lets ChatGPT list, search, and run your Apple Shortcuts via a local Mac agent
Use your own Mac from ChatGPT, Claude or Codex: files, commands, documents, and a browser.
The Remote MCP server acts as a standardized bridge between LLM applications (like Claude, ChatGPT, and Cursor) and external services, enabling AI agents to access external tools and resources. Its primary capability is providing a centralized search tool to discover other MCP servers and their respective tools. Unlike local implementations, it runs remotely with OAuth authentication and permission controls for security.
Connect any AI agent to 1,000+ apps and 27,000+ actions through one remote MCP server (OAuth).
Related MCP Servers
- AlicenseNot gradedqualityAmaintenanceEnables ChatGPT to access approved local Windows folders, run commands, control the desktop, and manage session history through a permission-bounded MCP server.4,202MIT
- AlicenseNot gradedqualityAmaintenanceEnables ChatGPT on Windows to securely access codebases, Git, terminals, language servers, debuggers, SQLite, local HTTP services, adaptive project memory, engineering skills, checkpoints, audit logs, and sandboxed command execution through MCP tools.MIT
- AlicenseNot gradedqualityBmaintenanceBridges the ChatGPT web UI to a local development environment over MCP, letting the model read and patch files, run real shell commands and terminals, execute tests, and control the desktop (screen, mouse, keyboard, clipboard) within user-approved folders. It adds parallel workers, goal/loop long-task control, plugin tools, and a security-hardening layer with risk-tiered shell policy, audit logging, and credential sanitization.MIT
- AlicenseNot gradedqualityBmaintenanceEnables ChatGPT to read and write local files, view the screen, and operate Windows desktop windows, mouse, and keyboard through MCP stdio and OpenAI's Secure MCP Tunnel.5MIT