Documentation
Everything from install to troubleshooting for MCP-CORE Personal — jump to a section below.
Getting started
-
1
Install MCP-CORE Personal
Download and run the Windows installer. The wizard asks for “just me” or “all users” and automatically prepares the required components (WebView2 and the .NET Desktop Runtime).
-
2
Open the app
The dashboard opens in its own desktop window (WebView2) — no separate browser needed. Closing the window keeps the app running in the tray; bring it back with “Show App” from the tray menu.
-
3
Follow the First-Run guide
On first launch, an ordered setup menu walks you through the steps needed to get the app fully working.
Runtimes
-
1
Open the Runtime Center
MCP servers need runtimes such as Node.js, Python or uv. Install or update them with one click from the Runtime Center.
-
2
The right version is chosen automatically
If an MCP requires a specific Node/Python version, a matching managed version is installed and used for that MCP; your system versions are left untouched.
-
3
Runtimes are kept in a shared location
Managed runtimes live in a shared machine folder (%PROGRAMDATA%\MCP-CORE Personal); they are not re-downloaded when the app updates or is reinstalled.
Installing MCP servers
-
1
Find one in the catalog
Use the MCP Discovery tab to search the catalog and filter by runtime and status to pick the server you need.
-
2
Review the requirements
Before installing, a preflight shows which runtime, environment variable or system tool is required; anything missing is clearly flagged.
-
3
Install and verify in one click
Fill in the install form fields (e.g. an API key) if required. The app installs the MCP, proves it works via initialize + tools/list, and shows the result.
-
4
Add a manual MCP
Add an MCP that is not in the catalog by command, package or URL; the app scans for the required parameters and presents them as a form.
Connecting AI apps
-
1
Supported apps
Claude Desktop, Claude Code, Cursor, VS Code and Codex are auto-configured with the connection helpers in the AI Applications tab.
-
2
Reversible changes
Connecting an app edits its config file; a backup is taken first and you can disconnect at any time.
-
3
The connection is actually verified
After the config is written, the app reports file validity, gateway reachability and whether a restart is needed — “written” does not mean “connected”.
Gateway & public URL
-
1
One local gateway URL
All installed MCPs are served through a single local gateway URL. Add this URL to your AI app; a local address is enough for desktop apps.
-
2
Public URL for remote clients
For clients that don’t support localhost (e.g. ChatGPT), you can open a keyed public URL. Only the MCP endpoint is exposed; the management API stays closed to public access.
-
3
Test the connection
The connection test in the Gateway tab verifies the URL really works via initialize and tools/list.
Monitoring tool calls
-
1
Live stream
Watch recent calls live in the Tool Calls tab; see each call’s status, duration and which MCP it was routed to.
-
2
Detail and export
Open a call’s detail to see the request/response summary, coerced arguments and a performance breakdown; export records as CSV or JSON.
Backup & restore
-
1
Create a backup
Back up your installed MCPs’ manifests, configs and AI client settings into a single archive.
-
2
Restore
Because archive paths are portable, you can restore a backup on the same or another installation.
Data & file locations
-
1
App data next to the install
MCPs, settings, logs and history are kept in the “data” folder next to the install (e.g. C:\Program Files\MCP-CORE Personal\data), so the install is self-contained.
-
2
Shared runtimes
Managed Node/Python/uv runtimes are shared under %PROGRAMDATA%\MCP-CORE Personal; they persist across app updates and are shared across users.
-
3
Move the data directory
From Settings you can move the MCP install/data directory to another drive; the app closes open files and reopens from the new location.
Updating & uninstalling
-
1
Update
A version check runs at startup. If an update is available, follow the notification to install the latest version.
-
2
Uninstall cleanly
If you choose “full removal” during uninstall, app data, shared runtimes and AI client connections are cleaned up too; “program only” keeps your data.
Troubleshooting
-
1
An MCP is not working
Check the health status and recommended actions on the Installed MCPs screen; open the logs or use “Repair” to cleanly reinstall the same version.
-
2
The AI app doesn’t see the MCPs
Most clients need a restart after a config change. Follow the restart note on the connection screen and verify the gateway URL.
-
3
Diagnostics bundle
From the Diagnostics tab, run a health check and collect logs and a diagnostics bundle to make an issue shareable.