SCALE API Helper MCP
Allows direct HTTP requests to the SCALE API proxy server using cURL.
Click on "Install 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., "@SCALE API Helper MCPlist all VMs"
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.
SCALE API Helper + MCP
A secure, LLM-ready bridge between your local AI environment and the SCALE Computing HyperCore REST API. Exposes VM management operations as authenticated FastAPI routes and MCP tools — usable from OpenWebUI, Claude Desktop, or cURL.
Built to go beyond API exploration: this is a functional operator tool for taking real actions against a SCALE HyperCore cluster.
Tested against SCALE HyperCore v9.4.30.217736 & v9.6.19.223352
Features
FastAPI proxy server with Bearer Auth securing all routes
MCP tool integration for OpenWebUI and Claude Desktop
Session-based auth against SCALE HyperCore (stored locally in
.pfile)VM lifecycle management: snapshot, clone, export, state changes
Swagger UI at
/docswith full OpenAPI schemaRAG-compatible schema querying for LLM-assisted API exploration
Related MCP server: proxmox-mcp
File Overview
File | Purpose |
| Route definitions, Bearer auth, OpenAPI schema, MCP forwarding |
| All SCALE API logic: session lifecycle, VM operations, query forwarding |
| Pydantic models for use with Swagger UI |
| Bearer token and cluster host configuration |
| Full HyperCore OpenAPI schema (used for RAG/exploration) |
| SQLite schema store for LLM-aided endpoint discovery |
| Setup and exploration scripts |
Requirements
Python 3.10+
MCP runtime via
uvx mcpoDependencies:
fastapi,httpx,pydantic,uvicornany OpenAI API Compat. Server — tested with
nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8on vLLM (0.18.1)
Setup
1. Clone & Install
git clone https://github.com/Nels2/SCALE-API-Helper-MCP
cd SCALE-API-Helper-MCP
uv pip install -r requirements.txt2. Configure
Edit config_alt.py:
MCP_BEARER_TOKEN = "your-secret-token"
SCALE_HOST = "your-cluster-ip"3. Run
uvx mcpo --port 5075 -- uvicorn 03_mcpserver_agent+auth:app --host 0.0.0.0 --port 5075Optionally, add --reload if you plan to make changes and dont wan't to kill the server.
Swagger UI: http://localhost:5075/docs
MCP Usage
OpenWebUI
uvx mcpo --port 5075 -- uv run fastapi_server.pyAdd connection in OpenWebUI:
URL: http://<your-ip>:5075/openapi.json
Auth: Bearer <your-token>Note: After adding new tools to the MCP server, restart the server process and reconnect in OpenWebUI — tool lists are cached at connection time.
Claude Desktop
Add to your claude_desktop_config.json:
{
"mcpServers": {
"scale-agent": {
"url": "http://<your-ip>:5075",
"headers": {
"Authorization": "Bearer <your-token>"
}
}
}
}Available Tools
Session Management
Tool | Description |
| Authenticate against SCALE and store session |
| Retrieve current session info (validates 12hr window) |
| Invalidate active session |
Read Operations
Tool | Description |
| Check connectivity and session validity |
| Retrieve cluster node information |
| Retrieve cluster spec/configuration |
| List all active conditions/alerts |
| List all users on the system |
| List all ISOs |
| List all physical drives |
| List all virtual disks |
| List all virtual network devices |
| List all VM snapshots |
| List all task/status tags |
| Get status for a specific task tag by ID |
VM Operations
Tool | Description |
| Search VMs by name or user (via NMAPX Dash) |
| Fetch full VM configuration by UUID |
| Retrieve live CPU, memory, disk, and network stats |
| Take action on a VM: |
| Take a named snapshot of a VM |
| Delete a snapshot by UUID |
| Clone a VM |
| Export a VM to the configured SMB archive target |
Raw Access
Tool | Description |
| Direct proxy call — any endpoint, method, and payload |
| Search the local schema DB for matching endpoints |
Architecture
OpenWebUI / Claude Desktop
│
▼
fastapi_server.py ←── Bearer Auth
│
▼
scale_tools.py ←── Session (.p file)
│
▼
SCALE HyperCore REST APINotes
This is a personal developer tool but it IS technically production-hardened since it is secured by Bearer Auth.
The SCALE API uses session cookies, not per-request tokens; the
.pfile persists this between calls, it is only valid based on the SCALE Permissions, but the script sets a hard time of 12 hours to auto refresh token.run_apigives you a raw escape hatch for any endpoint not yet wrapped as a named tool... but your model has to be smart enough to handle embedded JSON within JSON.
License
MIT. Use and adapt freely.
This server cannot be installed
Maintenance
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/Nels2/SCALE-API-Helper-MCP'
If you have feedback or need assistance with the MCP directory API, please join our Discord server