Model Manager runs on your own machine — the client starts it, so there is no endpoint to ping. 240 installs a week from pypi. Last commit 22 Oct 2026.
Manage your local model machine over MCP: state, model pull/remove, service switch, LoRA training
What the code does
We read the source, 7 d ago · rules 995398eba69f
What this server is able to do. For an MCP server this is often the job itself — a terminal server runs commands because that is what it is for. Listed so you know what you are plugging in, not as an accusation.
result = subprocess.run(
env = dict(os.environ)
TARGET="${MM_AUTHORIZED_KEYS:-$HOME/.ssh/authorized_keys}"
Is this your server and something here is wrong? Tell us — corrections are free and do not require a plan.
This code can reach further than it looks
We found places where it runs commands, builds paths or queries from values it is given. None of that is a flaw by itself — it becomes one when the code changes, and code changes quietly between releases. We re-read it on every one.
Connect this server
This server runs on your own machine — install it with the package manager and the client starts it for you. Package name taken from the official registry entry.
claude mcp add mcp-modelmanager -e 'MM_VM_HOST=<MM_VM_HOST>' -e 'MM_VM_USER=<MM_VM_USER>' -e 'MM_CONTAINER_ROOT=<MM_CONTAINER_ROOT>' -- uvx mcp-modelmanager{
"mcpServers": {
"mcp-modelmanager": {
"command": "uvx",
"args": [
"mcp-modelmanager"
],
"env": {
"MM_CONTAINER_ROOT": "<MM_CONTAINER_ROOT>",
"MM_VM_HOST": "<MM_VM_HOST>",
"MM_VM_USER": "<MM_VM_USER>"
}
}
}
}[mcp_servers.mcp-modelmanager]
command = "uvx"
args = ["mcp-modelmanager"]
env = { MM_VM_HOST = "<MM_VM_HOST>", MM_VM_USER = "<MM_VM_USER>", MM_CONTAINER_ROOT = "<MM_CONTAINER_ROOT>" }{
"mcpServers": {
"mcp-modelmanager": {
"command": "uvx",
"args": [
"mcp-modelmanager"
],
"env": {
"MM_CONTAINER_ROOT": "<MM_CONTAINER_ROOT>",
"MM_VM_HOST": "<MM_VM_HOST>",
"MM_VM_USER": "<MM_VM_USER>"
}
}
}
}{
"servers": {
"mcp-modelmanager": {
"type": "stdio",
"command": "uvx",
"args": [
"mcp-modelmanager"
],
"env": {
"MM_CONTAINER_ROOT": "<MM_CONTAINER_ROOT>",
"MM_VM_HOST": "<MM_VM_HOST>",
"MM_VM_USER": "<MM_VM_USER>"
}
}
}
}Put your own values in place of
<MM_VM_HOST> (Address of the model machine (127.0.0.1, LAN IP or tunnel/VPN address)), <MM_VM_USER> (SSH user on the model machine), <MM_CONTAINER_ROOT> (Root directory on the machine holding all model, adapter and cache data).
The author’s registry entry marks them as required.
Optional settings from the registry entry, left out of the block above:
MM_MAX_TIER (Highest tool tier exposed: read, control or destructive), MM_ENABLED_OPERATIONS (Optional comma-separated allowlist of exact tool names, applied on top of MM_MAX_TIER; a typo aborts startup. Empty means everything within MM_MAX_TIER), MM_VM2_HOST (Optional: address of a second machine sharing the GPU, used to interpret unreachability. Leave empty if there is only one machine).
Alternatives to Model Manager
same job, measured the same wayLocal-first AI memory MCP server. Record, recall and consolidate memory on your own machine.
Remote MCP server for Web3TV creators — manage your account over MCP.
Public links for local services and static artifacts: live from your machine or hosted by Serve.
Unofficial read-only MCP server for your local WhatsApp chat history (stays on your machine).
Local Kokoro-82M TTS MCP server that synthesizes and plays speech on your machine
Local-first MCP server for your own LINE and Telegram chats. Data stays on your machine.
Local, privacy-preserving PII toolkit over MCP — shape to the model, substance stays local.
Local-first MCP server that snapshots your project's working state into one context object.
Model Manager — questions
Answers built from our own checks of this server.