RentServers.com

Measured conclusion: Deep-tested. L2 outcome: tools_ok.

GPU and bare-metal price index, 36-month TCO scorecard, and blind counter-quotes.

Add to your agent

Copy this into your MCP client config:

{
  "mcpServers": {
    "RentServers.com": {
      "url": "https://rentservers.com/mcp",
      "transport": "streamable-http"
    }
  }
}

Measured tools (5)

• get_server_price_index read-only

  Current, dated rental price benchmarks. GPUs: on-demand and 12-month reserved median USD per GPU-hour, lowest on-demand listing, reserved price by term, and RentServers.com's published quote medians. Dedicated: entry floors, reference tiers and Hetzner/OVHcloud SKU prices. VPS: cheapest and median by size.

  Parameters:

  • model — string — GPU (h100, h200, b200, gb200, a100, l40s) or server (ax102, ax42, advance-2, scale-a1, game-1).
  • region — string — Filters published quote medians.
  • category — string — Defaults to gpu, or dedicated when model is a server SKU.

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• calculate_36_month_tco read-only

  Normalizes server costs over 36 months (or horizon_months): rent with announced price changes and an assumed yearly rise, setup fees, extra IPv4, egress overage, add-ons. Either pass incumbent_sku (an id from audit_2026_provider_repricing, e.g. ovh-advance-2-2026 or hetzner-ax102-1) with quantity, or pass scorecard: a full input with several quotes to compare (role incumbent|bid, units, pricing{base, unit month|gpu_hour, gpusPerUnit, promo}, setupFeePerUnit, extraIPv4, egress, sla, deliveryDays). Anything not stated is flagged and counted as $0.

  Parameters:

  • eur_usd — number — USD per EUR; defaults to the latest ECB reference rate
  • quantity — integer
  • scorecard — object — Full ScorecardInput, as used by rentservers.com/tools/36-month-tco/
  • extra_ipv4 — integer
  • incumbent_sku — string
  • horizon_months — integer
  • overage_usd_per_tb — number
  • egress_tb_per_month — number
  • annual_escalation_pct — number
  • ipv4_monthly_each_usd — number
  • included_tb_per_server — number

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• audit_2026_provider_repricing read-only

  Hetzner (April 1 and June 15, 2026) and OVHcloud (July/August option and configuration changes, renewals from October 1, 2026) price changes by SKU, with renewal rules and when to stay versus switch. Use the returned SKU ids with calculate_36_month_tco.

  Parameters:

  • sku — string — Optional, e.g. advance-2, scale-a1, ax102
  • provider — string

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

• request_blind_counter_quotes

  Opens a counter-quote request with the RentServers.com desk: 3+ providers bid blind against the spec and the reply lays every quote out on one 36-month basis. Only call this when the user asks for quotes and provides their work email; they must click a confirmation link before any provider is contacted. Requests under $100/mo or for a single VPS get self-serve list prices instead of a ticket.

  Parameters:

  • work_email (required) — string
  • category (required) — string
  • region (required) — string
  • name — string
  • notes — string
  • company — string
  • quantity — integer
  • gpu_count — integer
  • gpu_model — string
  • term_months — integer
  • renewal_date — string
  • spec_summary — string — e.g. 64 × H200 with InfiniBand, or 10 × 16-core/128 GB NVMe servers
  • incumbent_sku — string
  • incumbent_provider — string
  • monthly_budget_usd — number
  • incumbent_monthly_usd — number

  Hints: read-only no · destructive no · idempotent no · open-world yes

• calculate_managed_gcp_ai read-only

  Prices Managed Inference on Google Cloud (Model A): GPU clusters and vLLM endpoints that run in the client's own GCP project. Google bills compute to the client at list price; Wells Software Consulting LLC bills a flat management retainer per node ($250 L4, $800 A100 80GB, $2,500 H100). For a model, peak requests per second, token split and commitment tier it returns every node shape with GCP compute, retainer and total per month, plus node counts with N+1 once throughput baselines are published (sizing_ready).

  Parameters:

  • model (required) — string
  • peak_rps (required) — number — Peak requests per second
  • tier — string — GCP pricing tier. cud_3y also notes DWS calendar block prices for H100. Default on_demand
  • input_tokens — integer — Prompt tokens per request; default 512
  • output_tokens — integer — Output tokens per request; default 256

  Hints: read-only yes · destructive not declared · idempotent not declared · open-world no

Publisher website

Measured by MCPYP probing. Untested fields are omitted, never estimated. Listing is always free; MCPYP never proxies calls or stores user tokens.