Skip to content
MCP server

ComputeNexa

Find, quote, launch and stop private GPU machines on ComputeNexa from an AI assistant

Runs locallyStatus and history
Install
Pricing
Paid
Platforms
CLI, API
Runs
Locally
Version
0.1.0
Listing updated
3 days ago

About ComputeNexa

ComputeNexa is a paid MCP server for cloud and infrastructure. It runs locally on your machine and is started by your AI client.

Spending money is gated by a quote. launch_machine, extend_machine and set_keep_running first return a price, for example a two-hour A6000 session at $0.72 an hour with unused time refunded, and only go ahead on a second call with confirm=true and the quoted total. If the price rose in between the call is refused, and any single action above $25 is refused unless you raise CNX_MCP_MAX_LAUNCH_USD.

Keys come in two kinds: Look only, for browsing GPUs, machines and spending, and Launch and stop. Tools listed include whoami, list_gpus (live prices and stock, cheapest first), list_apps, list_machines, get_usage, list_receipts, stop_machine, and create_endpoint for a stable HTTPS address.

The server runs locally via uvx from PyPI, so your key stays in your own settings and never appears in tool output. ComputeNexa itself is in early access, with GPUs billed by the minute from prepaid credits.

Install

Run in a terminal

claude mcp add computenexa-mcp --env CNX_API_KEY='<CNX_API_KEY>' -- uvx computenexa-mcp

Add to claude_desktop_config.json

{
  "mcpServers": {
    "computenexa-mcp": {
      "command": "uvx",
      "args": [
        "computenexa-mcp"
      ],
      "env": {
        "CNX_API_KEY": "<CNX_API_KEY>"
      }
    }
  }
}

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "computenexa-mcp": {
      "command": "uvx",
      "args": [
        "computenexa-mcp"
      ],
      "env": {
        "CNX_API_KEY": "<CNX_API_KEY>"
      }
    }
  }
}

Add to .vscode/mcp.json

{
  "servers": {
    "computenexa-mcp": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "computenexa-mcp"
      ],
      "env": {
        "CNX_API_KEY": "<CNX_API_KEY>"
      }
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "computenexa-mcp": {
      "command": "uvx",
      "args": [
        "computenexa-mcp"
      ],
      "env": {
        "CNX_API_KEY": "<CNX_API_KEY>"
      }
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "computenexa-mcp": {
      "command": "uvx",
      "args": [
        "computenexa-mcp"
      ],
      "env": {
        "CNX_API_KEY": "<CNX_API_KEY>"
      }
    }
  }
}

Configuration

  • CNX_API_KEYA ComputeNexa API key (dashboard, API keys): Look only, or Launch and stopRequired
  • CNX_MCP_MAX_LAUNCH_USDMost one launch, extension or renewal may cost, in US dollars

Tools this server provides (15)

Taken from the project documentation; the running server may differ.

whoami
Your email, team, balance, GPU limit and key permissions.
list_gpus
GPUs with live prices and stock, cheapest first.
list_apps
Apps a GPU starts with, their models and what is installed.
list_machines
Your machines.
get_machine
How to connect to a machine.
get_usage
Spending and GPU-hours.
list_receipts
Finished sessions and their cost.
list_projects
Saved projects.
list_endpoints
Endpoints and the machine each points at.
launch_machine
Launches a GPU after a quote and confirmation.
extend_machine
Adds time after a quote.
set_keep_running
Renews a machine automatically until stopped.
stop_machine
Ends a machine now; unused time is refunded.
create_endpoint
Creates a stable HTTPS address for your own server.
get_endpoint_key
Returns an endpoint's key.

What it does

Quote before charge
Launch, extend and renew calls return a price and need confirmation.
Spend cap per call
Actions over $25 are refused by default; the limit is configurable.
Live GPU catalogue
Lists GPUs by price and stock, filterable by app, region and memory.
Look-only keys
Read-only keys expose only browsing and usage tools.
Usage and receipts
Shows spend, GPU-hours and finished sessions.
Stable endpoints
Create an HTTPS address pointing to your own server.

Good for

  • Ask which GPUs with 48 GB or more are in stock
  • Launch the cheapest machine with a data science app for two hours
  • Stop a machine and get unused time refunded
  • Check what the team spent this week

Pricing

Paid

Billed by the minute from prepaid credits

Prices change. Check computenexa.com before you buy.

ComputeNexa FAQ

Can the assistant overspend?

No. Money-spending tools return a quote first, and calls over $25 are refused by default.

How do I install it?

Install uv, create an API key in the dashboard, then run uvx computenexa-mcp with CNX_API_KEY set.

How is compute billed?

By the minute from prepaid credits; stopping early refunds unused time.

Is ComputeNexa free?

No. ComputeNexa is a paid product.

How do I add ComputeNexa to Claude Code?

Run this command in your terminal: claude mcp add computenexa-mcp --env CNX_API_KEY='<CNX_API_KEY>' -- uvx computenexa-mcp

What are the best alternatives to ComputeNexa?

The closest alternatives to ComputeNexa are Claude Code Ultimate Guide, Agent Yurt, Proofline, signafAIde and Toolkien.

Alternatives to ComputeNexa

  1. MCP server for searching the Claude Code Ultimate Guide from any MCP client

    Runs locally
    • Open source
    • 6.5K GitHub stars
    • AI models and LLM tooling
  2. Agent Yurt

    MCP server

    Back up agent memory, files and dashboards, then recover from a reviewed snapshot

    Not checked yet
    • Remote
    • Free
    • Knowledge and memory
  3. Proofline

    MCP server

    Production-aware review of changes against your AWS or Google Cloud setup, via MCP

    • Remote
    • Freemium
    • Developer tools
  4. signafAIde

    MCP server

    Local MCP server to set up, verify and check coverage of signed agent action records

    Runs locally
    • 5 tools
    • Freemium
    • Security
  5. Toolkien

    MCP server

    One MCP gateway for all your MCP servers, with code mode, permissions and audit

    Not checked yet
    • 3 tools
    • Remote
    • Open source
    • AI models and LLM tooling
  6. filex

    MCP server

    Self-hosted file manager in one Go binary with a built-in MCP server for agents

    Not checked yet
    • Remote
    • Open source
    • Files and documents