MCP SERVERWhat an H100 hour costs today, as a published daily median with a citation ready to paste.
LABSCO SUMMARY
GPU and token pricing is scattered across provider pages that each define things differently. Pulse publishes a daily median per index with the methodology version attached and a ready-made citation string, so the number in your model arrives with its provenance.
The MCP interface to Pulse, an open compute-pricing reference. It publishes daily medians for GPU rental and inference tokens under a CC-BY licence: four GPU series comparing hyperscaler against neocloud, four further neocloud series, and an inference-token basket anchored on an open Llama model at FP8.
Nothing — the server is hosted at https://pulsebenchmarks.com/mcp, JSON-RPC 2.0 over POST, no auth. Clients without remote support reach it through mcp-remote.
Paste a URL — add the endpoint to your client
USE CASES
Take the latest value for the index that matches your hardware, with the citation to put in the document.
Compare the indices side by side instead of reading two pricing pages that count differently.
Pull the series over 30 days, 90 days, a year or the whole history, with the per-provider observations underneath.
Read pipeline freshness and per-provider collection health, so a stale figure is caught rather than quoted.
List everything published with its current value and methodology version, before picking the one your question needs.
Counted once with only startup setup. Other versions, modes or optional credentials may expose a different list.
compare_indicesCompare current values across multiple Pulse indices. Use this when the user wants to see, for example, hyperscaler vs neocloud pricing for the same GPU, or one model vs another in the inference token index. Pass an array of slugs; returns a filtered summary.
get_seriesFetch the full payload for one Pulse index: metadata, the gated daily headline series, every per-provider observation behind it, and the latest contributing-provider list. Use this when the user wants history, distribution shape (P25/P75), or the underlying provider breakdown for a specific series. Optional range filter limits to the last N days.
get_statusPipeline freshness and per-provider collection health. Use this when the user asks 'is Pulse up-to-date?' or before quoting a number to confirm the pipeline isn't degraded. Returns overall status plus per-provider and per-index breakdowns.
latest_valueQuick lookup: the latest published value for one index, with minimal payload. Use this when the user asks 'what's the price of {GPU} on {provider type}' or 'what's the current Pulse {series} number'. Always quote the as-of date and the source family.
list_indicesList every published Pulse compute-pricing index with its current value, freshness, methodology version, and status. Use this when the user asks 'what does Pulse track?' or 'what's the current price of GPU compute?' Returns an array; quote the value alongside its source family (hyperscaler / neocloud) and as-of date.
BEFORE YOU INSTALL
5 tools when we connected. All five only read. get_series accepts a range of 30d, 90d, 1y or all, and latest_value returns a cite_as string for quoting a figure.
npx on PATH
All 4 documented clients connect to a hosted endpoint.
CONTEXT COST
Rated by the size of the 5 tool definitions that ride along on every turn, not by how many there are
INSTALLATION
Add the endpoint to your client.
https://pulsebenchmarks.com/mcp{
"mcpServers": {
"pulse": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://pulsebenchmarks.com/mcp"
]
}
}
}MCP servers
anshumax
prmail
parlay-run
v587d