Manage your servers directly from LM Studio using your local models. One click installs the ManageLM MCP server into LM Studio: no editing JSON, no terminal commands.
Click the Add to LM Studio button above. LM Studio will open and ask you to confirm adding the ManageLM MCP server. That's it. You're done.
Click Add to LM Studio above (make sure LM Studio is installed and running)
Confirm the MCP server in LM Studio when prompted
Open a chat with a tool-capable model (e.g. Qwen, Llama 3.1, or any model with function-calling). The ManageLM tools appear automatically.
First tool call opens your browser for OAuth login: approve, and LM Studio is connected
Self-hosted users: The one-click button points at app.managelm.com. For your own portal, use the manual config below and replace the URL.
If the deeplink doesn't work, or you're on a self-hosted portal, add the MCP server manually:
In LM Studio, open the Program tab in the right-hand sidebar
Click Install → Edit mcp.json
Paste the snippet below and save
{
"mcpServers": {
"managelm": {
"url": "https://app.managelm.com/mcp"
}
}
}
Developer Mode: Editing mcp.json requires LM Studio's Developer Mode. Toggle it from the bottom-left user menu.
Once connected, any tool-capable local model in LM Studio can:
"all"All requests go to your ManageLM portal, which dispatches to your agents. Your local model stays on your machine. The portal just exposes tools over MCP.
ManageLM uses OAuth 2.0 with PKCE. The first tool call triggers a browser authorization flow. Approve access once and tokens refresh automatically.
No account yet? Create one, free for up to 10 agents.
Safari on macOS: Safari's Intelligent Tracking Prevention blocks the OAuth redirect back to LM Studio's loopback callback (http://127.0.0.1:*). Use Chrome or Firefox as your default browser for the sign-in step, then switch back to Safari after. You only need to do this once per auth session; tokens refresh automatically afterward.
Make sure LM Studio is installed and the deeplink handler is registered. On first install, open LM Studio once so the OS registers the lmstudio:// protocol, then retry. If it still fails, use the manual config.
Your model may not support tool calls. Load a model tagged with tool use or function calling in the LM Studio model catalog.
Open your ManageLM portal, go to Settings > MCP Credentials, and make sure MCP is enabled for your account.
Some models trigger the tool silently on boot. Send a plain message like "list my servers" in LM Studio to force the tool call and open the browser.