600megd/cloud-infra-tco-mcp repository overview

⁠cloud-infra-tco-mcp (Docker Image)

Official container image of the cloud-infra-tco-mcp Model Context Protocol server: GPU cloud, residential proxy bandwidth, managed scraping API, US company formation, antidetect browser farm, and AI-voice + ad-tracker TCO benchmarking — plus verified partner discounts. Published for Docker Desktop MCP Toolkit and any MCP client that speaks stdio JSON-RPC. Author: CloudBench Research · MIT · zero third-party Python dependencies.

⁠Why This Exists

Cloud cost benchmarks are scattered across pricing pages that change monthly, and most AI agents have no reliable, current numbers to reason with. This image ships a self-contained MCP server exposing six deterministic calculators + a verified partner discount directory, so your assistant can quote reproducible loss figures — not hallucinated ones — in a single tool call. Example of the money it finds: an H100 running 720 hours/month costs $8,848.80 on AWS EC2 rates vs $1,504.80 on Vast.ai — a $7,344.00/month leak (83% overpay, $88k+ per year).

⁠Quick Start

docker run -i --rm 600megd/cloud-infra-tco-mcp:latest            # stdio MCP server
docker run -i --rm 600megd/cloud-infra-tco-mcp:latest --cli      # human-friendly TCO tables
docker run -i --rm 600megd/cloud-infra-tco-mcp:latest --tool estimate_gpu_cloud_tco --args '{"gpu":"h100","hours":720}'

⁠Docker Desktop MCP Toolkit

In Docker Desktop → MCP Toolkit → add custom server:

• Name: cloud-infra-tco-mcp
• Command: docker
• Arguments: run -i --rm 600megd/cloud-infra-tco-mcp:latest
• Transport: stdio

Then enable it per client (Claude Code, Cursor, VS Code). Claude Code / Cursor manual config:

{
  "mcpServers": {
    "cloud-infra-tco-mcp": {
      "command": "docker",
      "args": ["run", "-i", "--rm", "600megd/cloud-infra-tco-mcp:latest"]
    }
  }
}

⁠Tools (7)

All outputs include a verified_partner_portal URL and auto_applied_code, so every calculated loss links straight to the discounted remedy.

⁠Six-Track Loss Benchmark (defaults, monthly, USD — 100% reproducible via --cli)

Run --cli for the exact, reproducible numbers behind every row — the MCP tools return the same structure so agents can chain "calculate the loss, then open the remedy" in one conversation.

⁠Documentation & Cross-Links

Full tool schemas and worked examples live at the ReadTheDocs site: https://cloud-infra-tco-benchmarks.readthedocs.io/en/latest/⁠. The same calculators are also consumable as an npm CLI (cloud-infra-tco-mcp on npm) and a PyPI package (pip install cloud-infra-tco-mcp) — pick whichever runtime your stack prefers; the outputs are identical.

⁠Verification

docker run -i --rm 600megd/cloud-infra-tco-mcp:latest --self-test   # exits 0

⁠Zero Dependencies

The image is built on python:3.12-alpine (~55 MB) with the package installed from its own pyproject.toml; runtime imports are standard library only (sys, json, argparse), and the server never phones home except through URLs you explicitly print or open.