InfraRabbit
Published list prices of compute (VPS, GPU hours, LLM tokens, media), archived daily. Read-only.
사용해야 할까요
품질 및 안전성
도구 정의와 프로토콜 준수에 대한 자동 분석을 기반으로 합니다.
컨텍스트 비용
이는 서버의 도구가 모델의 컨텍스트에 로드될 때마다 소비되는 대략적인 토큰 수입니다. 수치가 높을수록 다른 작업에 사용할 수 있는 주의가 줄어듭니다.
설치
원클릭 설치
`claude_desktop_config.json` 파일에 다음을 추가하세요:
{
"mcpServers": {
"data": {
"url": "https://infrarabbit.com/mcp"
}
}
}원격 엔드포인트
https://infrarabbit.com/mcpstreamable-http할 수 있는 일
도구 목록
도구 (13)
🟢search_skus(query, category, provider, limit, listing, ...)
Find VPS plans, GPU instances or LLM models by words ('openai gpt-5', 'H200', 'official anthropic', 'gemini 3'): every word must be in the SKU id or name; 'official' and lab names are signals. Rows come first-party first (the provider's own price, source_class first_party), then resellers (Amazon Bedrock, Azure ...), then community catalogues (models.dev, LiteLLM, OpenRouter), each row labelled with source_class. gpu and vps rows follow `commitment` (default on-demand: spot and committed prices are left out, as in cheapest_for_spec). Answers with a short `answer` sentence first (human names, price, VAT basis, term, read date, source), then compact rows; verbose=true returns every field of every row. No hit: near matches and did_you_mean. Active SKUs on the provider's public price page or product API by default (listing filter). Read-only; reads the ledger's own daily archive, nothing is fetched live.
입력 스키마
{
"type": "object",
"properties": {
"query": {
"default": "",
"description": "Words of the SKU name or id, e.g. 'h100', 'openai gpt-5', 'official anthropic'. Empty matches every active SKU, first-party first.",
"title": "Query",
"type": "string"
},
"category": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"description": "vps, gpu or llm",
"title": "Category"
},
"provider": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"description": "Provider slug, e.g. vultr, ovhcloud, openai",
"title": "Provider"
},
"limit": {
"default": 25,
"description": "Maximum rows, 1 to 200",
"title": "Limit",
"type": "integer"
},
"listing": {
"default": "current",
"description": "Listing filter. current (default) = on the provider's public price page or product API; catalog_only = only in an order catalog, not on the public page; vanished = was listed, now missing; unknown = no public page to check against; all = everything. Every row carries listing_status.",
"enum": [
"current",
"all",
"listed",
"catalog_only",
"vanished",
"unknown"
],
"title": "Listing",
"type": "string"
},
"commitment": {
"default": "on-demand",
"description": "gpu and vps rows only. Default on-demand: standing list prices with no term and no interruption. spot, reserved, commit, dynamic or community select those rows; any shows every commitment.",
"enum": [
"on-demand",
"spot",
"reserved",
"commit",
"dynamic",
"community",
"any"
],
"title": "Commitment",
"type": "string"
},
"verbose": {
"default": false,
"description": "true = every field of every row (spec, all price points, extra, cite) and the coverage gaps",
"title": "Verbose",
"type": "boolean"
}
},
"title": "search_skusArguments"
}🟢get_price(sku_id)
Current list price of one SKU (id like vps/ovhcloud/vps-2027-model2 or llm/official/anthropic/claude-sonnet-5-5) with all secondary prices, USD/EUR equivalents, spec, source URL, verified_at, a stale flag that is true when the source's latest fetch failed (the price shown is then the last verified one), other listings of the same model, and headline_commit where the provider headlines a commitment price (a labelled secondary price; the primary price has no commitment). Returns an error object for an unknown id.
입력 스키마
{
"type": "object",
"properties": {
"sku_id": {
"description": "SKU id as returned by search_skus, e.g. vps/ovhcloud/vps-2027-model2",
"title": "Sku Id",
"type": "string"
}
},
"required": [
"sku_id"
],
"title": "get_priceArguments"
}🟢get_history(sku_id, days)
Price change points of a SKU over the last N days (default 90): each row is when a price started to hold, with USD/EUR and the FX rate used. Rows are change points, not daily samples. Returns an error object for an unknown id.
입력 스키마
{
"type": "object",
"properties": {
"sku_id": {
"description": "SKU id as returned by search_skus",
"title": "Sku Id",
"type": "string"
},
"days": {
"default": 90,
"description": "Look-back window in days, 1 to 90 (the free window; longer per-SKU history is part of the paid API plans)",
"title": "Days",
"type": "integer"
}
},
"required": [
"sku_id"
],
"title": "get_historyArguments"
}🟢list_changes(since, category, provider, kind, limit)
Dated price changes, new SKUs, delistings and spec changes, newest first. Each row says when the change was first seen and the previous good snapshot it is measured against, plus the source id. Filter by since (ISO date), kind, category and provider.
입력 스키마
{
"type": "object",
"properties": {
"since": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"description": "ISO date or datetime, e.g. 2026-10-01",
"title": "Since"
},
"category": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"description": "vps, gpu or llm",
"title": "Category"
},
"provider": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"description": "Provider slug, e.g. vultr",
"title": "Provider"
},
"kind": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"description": "price_change, new_sku, delisted or spec_change",
"title": "Kind"
},
"limit": {
"default": 50,
"description": "Maximum rows, 1 to 500",
"title": "Limit",
"type": "integer"
}
},
"title": "list_changesArguments"
}🟢changes_since(since, kind, category, provider, limit, ...)
Everything recorded since a date, for a model whose knowledge ends at a cutoff: price changes, new items, removals and retirements, newest first, each with its UTC date, the previous good snapshot, the source URL and a permanent link to the item. counts are for the whole period (not just the page); gaps names the sources that are stopped now, with the day they stopped (never filled in; outages that recovered are not recorded); record_begins says how far back the record goes. summary=true returns the largest percent moves per category (at most 40 entries) in one answer, ordered by size of the move, not by price. Paging: pass next_cursor back as cursor. Entries carry recorded values only, never a current price. Returns an error object for a bad date or filter. Same data as GET /v1/changes-since. Use this one to page through every entry or to filter by category and kind (price_change, new_item, removed_or_stopped, retirement); for a first look at what changed since your cutoff, with counts, scope llm/gpu/vps and the gaps, call what_changed_since instead.
입력 스키마
{
"type": "object",
"properties": {
"since": {
"description": "UTC date, YYYY-MM-DD, e.g. the day your knowledge ends",
"title": "Since",
"type": "string"
},
"kind": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"description": "price_change, new_item, removed_or_stopped or retirement",
"title": "Kind"
},
"category": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"description": "vps, gpu, llm, inference or media",
"title": "Category"
},
"provider": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"description": "Provider slug, e.g. vultr, openai",
"title": "Provider"
},
"limit": {
"default": 200,
"description": "Maximum entries per page, 1 to 1000",
"title": "Limit",
"type": "integer"
},
"cursor": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"description": "next_cursor of the previous page",
"title": "Cursor"
},
"summary": {
"default": false,
"description": "true = the largest moves per category in one answer",
"title": "Summary",
"type": "boolean"
}
},
"required": [
"since"
],
"title": "changes_sinceArguments"
}🟢what_changed_since(date, kind, provider, limit)
Call this when a user asks about current prices, models or retirements and your knowledge may be out of date: give the date your knowledge ends (YYYY-MM-DD) and get everything recorded since, grouped: new models and plans, price changes (old to new, percent, date first seen, source URL), retirements a provider's own lifecycle page dates (effective, and scheduled ones still ahead), and removed plans. kind=llm (LLM API models and inference hosts), gpu, vps or all. counts are for the whole period; each group lists at most `limit` entries and, when cut, says so and links the full CSV; summary is a one-line answer. gaps names the days with no data and the sources that are stopped now (never filled in); record_begins and notice say how far back the record goes. Entries carry recorded values only, never a current price: use search_skus or get_price for today's price. Returns an error object for a bad or future date. Retirements come from Azure VM and AWS Bedrock lifecycle pages only (other providers are not parsed). Same answer as GET /v1/since and the page /prices/since/<date>. To page through every entry or filter by category use changes_since. Robots.txt changes of AI crawlers are a separate dataset (related.robots_txt_policy_changes, tool crawlers_changes).
입력 스키마
{
"type": "object",
"properties": {
"date": {
"description": "UTC date, YYYY-MM-DD: the day your knowledge ends, e.g. 2026-06-01",
"title": "Date",
"type": "string"
},
"kind": {
"default": "all",
"description": "all (default), llm (LLM API models and inference hosts), gpu or vps",
"enum": [
"all",
"llm",
"gpu",
"vps"
],
"title": "Kind",
"type": "string"
},
"provider": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"description": "Provider slug, e.g. openai, hetzner",
"title": "Provider"
},
"limit": {
"default": 25,
"description": "Entries per group, 1 to 200",
"title": "Limit",
"type": "integer"
}
},
"required": [
"date"
],
"title": "what_changed_sinceArguments"
}🟢get_uptime(host, days)
Uptime and latency of ~24 public AI API hosts (OpenAI, Anthropic, Google Gemini, Mistral, xAI, DeepSeek, Groq, Together, Fireworks, OpenRouter, Cohere, ElevenLabs, AWS Bedrock and more), measured by InfraRabbit itself every 5 minutes, not reported by the provider. Without `host`: every host with its current state and uptime %, p50 and p95 latency for the last 24 h, 7 d and 30 d. With `host` (the hostname, e.g. api.openai.com): its daily rows for `days` days and its incidents (two or more consecutive down windows, with start, end and evidence). Method: one unauthenticated HTTPS GET per host per round (DNS, TCP, TLS, time to first byte, HTTP status), from one vantage. It proves the provider's edge or gateway answered from there; it does NOT prove that a model generates or that your key, quota or region works. Check coverage_pct before quoting a figure: it says how much of the window was measured. Returns an error object for an unknown host. Same data as GET /v1/uptime and /v1/uptime/{host}/history and the pages /prices/uptime (status right now, worst first) and /prices/uptime/<provider>-api ("Is the OpenAI API down?").
입력 스키마
{
"type": "object",
"properties": {
"host": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"description": "Hostname as listed by this tool without a host, e.g. api.openai.com. Omit for every host.",
"title": "Host"
},
"days": {
"default": 30,
"description": "Days of daily rows and incidents when a host is given, 1 to 365",
"title": "Days",
"type": "integer"
}
},
"title": "get_uptimeArguments"
}🟢cheapest_for_spec(category, vcpu, ram_gb, disk_gb, gpu, ...)
Cheapest active SKUs meeting a minimum spec, ranked by list price only (no sponsorship, no affiliate weighting), excl. VAT (VAT-inclusive plans are ranked after removing their stated VAT rate). Starts with a short `answer` sentence (human names, price excl. VAT in the row's own currency, term, read date, source), then compact rows; each row carries price_excl_vat, source_class and, for GPUs, gpu_count and price_per_gpu_hour. Only first-party prices rank (the provider's own page or API); resellers, community catalogues (models.dev, LiteLLM, OpenRouter) and marketplaces take part only with include_community=true. gpu and vps default to commitment=on-demand: spot, reserved, committed, dynamic and community-cloud prices are left out unless commitment says otherwise. category vps|gpu|llm; vcpu, ram_gb, disk_gb minimums; gpu model substring (e.g. H100; with a memory size such as A100-80GB the card must state exactly that memory); region 'eu' for EU-deployable; currency USD|EUR for the ranking (rank.value is in that currency, rank.native_value in the row's own). verbose=true adds the full ranking method, every row field and the coverage gaps.
입력 스키마
{
"type": "object",
"properties": {
"category": {
"default": "vps",
"description": "vps, gpu or llm",
"title": "Category",
"type": "string"
},
"vcpu": {
"anyOf": [
{
"type": "number"
},
{
"type": "null"
}
],
"default": null,
"description": "Minimum vCPUs",
"title": "Vcpu"
},
"ram_gb": {
"anyOf": [
{
"type": "number"
},
{
"type": "null"
}
],
"default": null,
"description": "Minimum RAM in GB",
"title": "Ram Gb"
},
"disk_gb": {
"anyOf": [
{
"type": "number"
},
{
"type": "null"
}
],
"default": null,
"description": "Minimum disk in GB",
"title": "Disk Gb"
},
"gpu": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"description": "GPU model substring, e.g. H100",
"title": "Gpu"
},
"gpu_count": {
"anyOf": [
{
"type": "integer"
},
{
"type": "null"
}
],
"default": null,
"description": "Minimum GPUs per instance",
"title": "Gpu Count"
},
"region": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"description": "'eu' for EU-deployable, 'us', 'asia', a country code such as nl or de (the places of the compare page), or a provider region code",
"title": "Region"
},
"provider": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"description": "Provider slug to restrict to",
"title": "Provider"
},
"currency": {
"default": "USD",
"description": "USD or EUR (ranking currency)",
"title": "Currency",
"type": "string"
},
"limit": {
"default": 10,
"description": "Maximum rows, 1 to 50",
"title": "Limit",
"type": "integer"
},
"listing": {
"default": "current",
"description": "Listing filter, default current: only plans on the provider's public price page or product API count. all includes catalog-only and vanished plans (each row carries listing_status).",
"enum": [
"current",
"all",
"listed",
"catalog_only",
"vanished",
"unknown"
],
"title": "Listing",
"type": "string"
},
"commitment": {
"default": "on-demand",
"description": "gpu and vps only. Default on-demand: standing list prices with no term and no interruption. spot, reserved, commit, dynamic or community select those rows; any shows every commitment (each row carries extra.commitment).",
"enum": [
"on-demand",
"spot",
"reserved",
"commit",
"dynamic",
"community",
"any"
],
"title": "Commitment",
"type": "string"
},
"include_stale": {
"default": false,
"description": "Also rank plans whose price is not verified (stale): after every current one, each flagged stale. Off by default.",
"title": "Include Stale",
"type": "boolean"
},
"include_region_unstated": {
"default": false,
"description": "vps with a region filter only. A plan whose source states no region is never guessed into the place: it is left out of the rows, counted in region_unstated (count, and the cheapest of them). 1 = rank those plans with the others (the compare page's region_unknown=1).",
"title": "Include Region Unstated",
"type": "boolean"
},
"include_community": {
"default": false,
"description": "Also rank resellers (Amazon Bedrock, Azure ...), community catalogues (models.dev, LiteLLM, OpenRouter) and marketplaces; each row says source_class. Off by default: only the provider's own list prices rank.",
"title": "Include Community",
"type": "boolean"
},
"verbose": {
"default": false,
"description": "true = the full ranking method, every field of every row and the coverage gaps",
"title": "Verbose",
"type": "boolean"
}
},
"title": "cheapest_for_specArguments"
}🟢vps_plans(vcpu, ram_gb, disk_gb, max_price, region, ...)
VPS plans as /prices/vps/plans and GET /v1/vps/plans return them: the very same function, so the answers cannot differ for the same filters. Every row is a plan with its verified list price a month (excl. VAT) for the chosen contract term, all price points the provider states, buying facts (IPv4, traffic, disk and vCPU type; null = not stated), region (region_stated=false when the source names none), verified_at and source_url. Only plans with a price verified in the last 3 days carry a price; the others are listed apart in not_verified without one. Sorted by price (or per GB RAM, per vCPU); nothing is ranked for money. Each plan has a detail_url with its facts sheet. Every plan carries price_excl_vat in the provider's own currency. Answers with a short `answer` first and compact plans; verbose=true returns everything. Answer pages for common searches: /prices/vps/cheapest.
입력 스키마
{
"type": "object",
"properties": {
"vcpu": {
"anyOf": [
{
"type": "number"
},
{
"type": "null"
}
],
"default": null,
"description": "Minimum vCPU",
"title": "Vcpu"
},
"ram_gb": {
"anyOf": [
{
"type": "number"
},
{
"type": "null"
}
],
"default": null,
"description": "Minimum RAM in GB",
"title": "Ram Gb"
},
"disk_gb": {
"anyOf": [
{
"type": "number"
},
{
"type": "null"
}
],
"default": null,
"description": "Minimum disk in GB",
"title": "Disk Gb"
},
"max_price": {
"anyOf": [
{
"type": "number"
},
{
"type": "null"
}
],
"default": null,
"description": "Maximum price a month in the chosen currency",
"title": "Max Price"
},
"region": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"description": "A place key: eu, us, asia, a country code such as de or nl, or none (plans whose source states no region)",
"title": "Region"
},
"provider": {
"anyOf": [
{
"items": {
"type": "string"
},
"type": "array"
},
{
"type": "null"
}
],
"default": null,
"description": "Provider slugs, e.g. hetzner",
"title": "Provider"
},
"sort": {
"default": "price",
"description": "Sort order",
"enum": [
"price",
"per_gb",
"per_vcpu"
],
"title": "Sort",
"type": "string"
},
"currency": {
"default": "eur",
"description": "eur (default) or usd",
"enum": [
"eur",
"usd"
],
"title": "Currency",
"type": "string"
},
"term": {
"default": "none",
"description": "Contract term: none = no commitment (default), 12 = up to 12 months, any",
"enum": [
"none",
"12",
"any"
],
"title": "Term",
"type": "string"
},
"ipv4": {
"default": false,
"description": "Only plans whose source states an own IPv4 address is included",
"title": "Ipv4",
"type": "boolean"
},
"dedicated": {
"default": false,
"description": "Only plans whose source states dedicated vCPUs",
"title": "Dedicated",
"type": "boolean"
},
"traffic_tb": {
"anyOf": [
{
"type": "number"
},
{
"type": "null"
}
],
"default": null,
"description": "Minimum included traffic in TB a month (unlimited counts)",
"title": "Traffic Tb"
},
"nvme": {
"default": false,
"description": "Only plans whose source states NVMe disks",
"title": "Nvme",
"type": "boolean"
},
"limit": {
"default": 25,
"description": "Maximum plans returned, 1 to 500 (the same rows as the API; the API default is 200)",
"title": "Limit",
"type": "integer"
},
"verbose": {
"default": false,
"description": "true = every plan with all price points and facts, cheapest_per_provider, not_verified and the notes; default = a short answer and compact plans",
"title": "Verbose",
"type": "boolean"
}
},
"title": "vps_plansArguments"
}🟡stack_quote(items, idempotency_key)
One call, the best current answer for a whole stack: a list of items (kind vps: vcpu, ram_gb, disk_gb, region, term, currency, count; kind gpu: gpu, count, hours_per_month, region, gpus_per_instance; kind llm or inference: model, input_mtok, output_mtok, count). Per item the 5 cheapest offers with a verified price (the free ordering: price only, nothing for money), the monthly cost for your count and usage (a stated unit price times your quantities, nothing estimated), verified_at, source_url and a scope line; an item with no fresh match says so. Totals per currency, never converted. Same answer as POST /v1/quote. Free: 3 quotes a day per address; beyond that 1 credit per quote on an API key (see /pricing). Reads prices, and records the free quote or credit it uses (so it is not read-only). Pass an idempotency_key to retry without a second charge. Returns an error object (HTTP 402 body) when out of credits.
입력 스키마
{
"type": "object",
"properties": {
"items": {
"description": "e.g. [{\"kind\":\"vps\",\"vcpu\":4,\"ram_gb\":8,\"region\":\"eu\",\"term\":\"any\",\"count\":3},{\"kind\":\"gpu\",\"gpu\":\"H100\",\"count\":1,\"hours_per_month\":200},{\"kind\":\"llm\",\"model\":\"gpt-5\",\"input_mtok\":50,\"output_mtok\":10}]",
"items": {
"additionalProperties": true,
"type": "object"
},
"title": "Items",
"type": "array"
},
"idempotency_key": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"description": "1 to 128 characters; the same key never charges twice (API key callers)",
"title": "Idempotency Key"
}
},
"required": [
"items"
],
"title": "stack_quoteArguments"
}🟢estimate_cost(sku_id, input_tokens, output_tokens, cached_input_tokens, hours, ...)
Itemised cost of a usage profile at the SKU's current list price: tokens (input/output/cached) for LLM SKUs, hours/months for instances. Returns line items with the rates used, native, USD and EUR totals, verified_at and source_url. Pure arithmetic on stored prices; excl. VAT unless the result says otherwise.
입력 스키마
{
"type": "object",
"properties": {
"sku_id": {
"description": "SKU id as returned by search_skus",
"title": "Sku Id",
"type": "string"
},
"input_tokens": {
"default": 0,
"description": "Input tokens (LLM SKUs)",
"title": "Input Tokens",
"type": "number"
},
"output_tokens": {
"default": 0,
"description": "Output tokens (LLM SKUs)",
"title": "Output Tokens",
"type": "number"
},
"cached_input_tokens": {
"default": 0,
"description": "Cached input tokens (LLM SKUs)",
"title": "Cached Input Tokens",
"type": "number"
},
"hours": {
"default": 0,
"description": "Instance hours (VPS/GPU SKUs)",
"title": "Hours",
"type": "number"
},
"months": {
"default": 0,
"description": "Instance months (VPS/GPU SKUs)",
"title": "Months",
"type": "number"
}
},
"required": [
"sku_id"
],
"title": "estimate_costArguments"
}🟢get_index(series, days)
A named daily index series (e.g. eu-vps-floor-2c4g, h100-list-median, frontier-input-floor) with its latest value, the number of constituents and the constituents themselves, so the value can be recomputed. Omit series to list all series with their latest values.
입력 스키마
{
"type": "object",
"properties": {
"series": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"description": "Series slug; omit to list all series",
"title": "Series"
},
"days": {
"default": 90,
"description": "Look-back window in days, 1 to 3650",
"title": "Days",
"type": "integer"
}
},
"title": "get_indexArguments"
}🟡alert_me(kind, target)
Where a person can get a free daily email when something changes: pass what to follow and get the subscribe URL to show them. kind = sku (a SKU id from search_skus: its price changes and retirement), provider (any change of a provider slug such as anthropic), category (vps, gpu, llm, inference or media), index (a series slug from get_index: moves of 5% or more) or new_model (new models of a provider). The person opens the URL, types their email address, confirms by link and gets at most one mail a day with a one-click unsubscribe. Nothing is sent from here and no address is taken by this tool. Returns an error object for an unknown target.
입력 스키마
{
"type": "object",
"properties": {
"kind": {
"description": "What kind of thing to follow",
"enum": [
"sku",
"provider",
"category",
"index",
"new_model"
],
"title": "Kind",
"type": "string"
},
"target": {
"description": "SKU id, provider slug, category, index series slug or provider slug (new_model)",
"title": "Target",
"type": "string"
}
},
"required": [
"kind",
"target"
],
"title": "alert_meArguments"
}커뮤니티
증거