LLM API token prices and GPU rental prices, each with its source and date. No key, no account.
Maintain this server? Claim it
Call this server's tools from your browser. Nothing to install.
Requests go through mcptry because browsers can't call most MCP servers directly. Header values are sent to the server for that request only and are never stored or logged.
Tool list from our last check. Connect live to refresh it.
Models ordered by their lowest input price on any channel, with the channel that offers it. Optional vendor or channel kind (OFFICIAL, CLOUD, AGGREGATOR). List prices as published by each channel at observed_at, USD per 1M tokens unless stated; China-site price lists are in CNY, listed separately and never converted. Standard (on-demand) plan unless service_plan says otherwise. The channel's own page is authoritative. Aggregator prices are the marketplace's own and never the reference. A missing price is unknown, not free. Not purchasing advice.
Pick your client. Values in capitals, like YOUR_TOKEN, are yours to fill in.
Run in a terminal
claude mcp add --transport http auramatrix https://auramatrix.ai/mcp
Models ordered by their lowest input price on any channel, with the channel that offers it. Optional vendor or channel kind (OFFICIAL, CLOUD, AGGREGATOR). List prices as published by each channel at observed_at, USD per 1M tokens unless stated; China-site price lists are in CNY, listed separately and never converted. Standard (on-demand) plan unless service_plan says otherwise. The channel's own page is authoritative. Aggregator prices are the marketplace's own and never the reference. A missing price is unknown, not free. Not purchasing advice.
| Argument | Type | Description |
|---|---|---|
| kind | string | |
| limit | integer | |
| vendor | string |
{
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"properties": {
"kind": {
"enum": [
"OFFICIAL",
"CLOUD",
"AGGREGATOR"
],
"type": "string"
},
"limit": {
"type": "integer",
"maximum": 50,
"minimum": 1
},
"vendor": {
"type": "string",
"maxLength": 40
}
},
"additionalProperties": false
}Places a price someone pays or was quoted among the prices on record. For a model API: where it stands against the official list price, the lowest recorded channel and the batch price. For a GPU: how many providers' on-demand list prices it undercuts, and the spot, reserved and marketplace prices. Use it for "is $X a good price for Y". Nothing is stored.
| Argument | Type | Description |
|---|---|---|
| item* | string | Model id or alias (token), or GPU id such as h100 (gpu) |
| kind* | string | |
| term | string | gpu only; default ON_DEMAND |
| price* | number | token: input price in USD per 1M tokens; gpu: price per GPU-hour |
| currency | string | gpu only; CNY compares with mainland China prices |
| output_price | number | token only: output price in USD per 1M tokens |
{
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"kind",
"item",
"price"
],
"properties": {
"item": {
"type": "string",
"maxLength": 120,
"minLength": 1,
"description": "Model id or alias (token), or GPU id such as h100 (gpu)"
},
"kind": {
"enum": [
"token",
"gpu"
],
"type": "string"
},
"term": {
"enum": [
"ON_DEMAND",
"SPOT",
"MONTH_1",
"YEAR_1",
"YEAR_3"
],
"type": "string",
"description": "gpu only; default ON_DEMAND"
},
"price": {
"type": "number",
"maximum": 1000000,
"description": "token: input price in USD per 1M tokens; gpu: price per GPU-hour",
"exclusiveMinimum": 0
},
"currency": {
"enum": [
"USD",
"CNY"
],
"type": "string",
"description": "gpu only; CNY compares with mainland China prices"
},
"output_price": {
"type": "number",
"maximum": 1000000,
"description": "token only: output price in USD per 1M tokens",
"exclusiveMinimum": 0
}
},
"additionalProperties": false
}All channels for a model sorted by input price, with the difference versus the official list price. List prices as published by each channel at observed_at, USD per 1M tokens unless stated; China-site price lists are in CNY, listed separately and never converted. Standard (on-demand) plan unless service_plan says otherwise. The channel's own page is authoritative. Aggregator prices are the marketplace's own and never the reference. A missing price is unknown, not free. Not purchasing advice.
| Argument | Type | Description |
|---|---|---|
| model* | string | Model id such as openai/gpt-6.1-sol, or an alias the channels use (gpt-6.1-sol, claude-opus-5-5). Case-insensitive. |
{
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"model"
],
"properties": {
"model": {
"type": "string",
"maxLength": 120,
"minLength": 1,
"description": "Model id such as openai/gpt-6.1-sol, or an alias the channels use (gpt-6.1-sol, claude-opus-5-5). Case-insensitive."
}
},
"additionalProperties": false
}Cost of one request from the recorded prices and the channel's billing rules (long-context thresholds, time-of-day prices, thinking-token billing), each rule with its source excerpt. Where the sources leave a question open the answer is a range with the reason, never a guess; a price that is not recorded is unknown. List prices as published by each channel at observed_at, USD per 1M tokens unless stated; China-site price lists are in CNY, listed separately and never converted. Standard (on-demand) plan unless service_plan says otherwise. The channel's own page is authoritative. Aggregator prices are the marketplace's own and never the reference. A missing price is unknown, not free. Not purchasing advice.
| Argument | Type | Description |
|---|---|---|
| at | string | Request time, ISO 8601, for channels that price by time of day |
| model* | string | Model id such as openai/gpt-6.1-sol, or an alias the channels use (gpt-6.1-sol, claude-opus-5-5). Case-insensitive. |
| channel | string | Channel id; default is the official channel |
| input_tokens* | integer | Uncached input tokens |
| service_plan | string | Default STANDARD |
| output_tokens* | integer | Visible output tokens, without thinking tokens |
Latest standard price per channel for one model, each with region, pricing mode, trust status, observation time and source snapshot, plus Batch / Flex / Priority / Ultrafast plan prices where a channel publishes them. List prices as published by each channel at observed_at, USD per 1M tokens unless stated; China-site price lists are in CNY, listed separately and never converted. Standard (on-demand) plan unless service_plan says otherwise. The channel's own page is authoritative. Aggregator prices are the marketplace's own and never the reference. A missing price is unknown, not free. Not purchasing advice.
| Argument | Type | Description |
|---|---|---|
| model* | string | Model id such as openai/gpt-6.1-sol, or an alias the channels use (gpt-6.1-sol, claude-opus-5-5). Case-insensitive. |
| channel | string | Channel id to restrict to, e.g. azure-datazone-eu, vertex-regional, openrouter |
{
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"model"
],
"properties": {
"model": {
"type": "string",
"maxLength": 120,
"minLength": 1,
"description": "Model id such as openai/gpt-6.1-sol, or an alias the channels use (gpt-6.1-sol, claude-opus-5-5). Case-insensitive."
},
"channel": {
"type": "string",
"maxLength": 80,
"description": "Channel id to restrict to, e.g. azure-datazone-eu, vertex-regional, openrouter"
}
},
"additionalProperties": false
}What renting a GPU costs per hour. Without arguments: every GPU model with its on-demand range across providers. With a gpu id: every provider, kind (on demand, spot, reserved, marketplace) and machine, cheapest first, each with its source. Prices in mainland China (Alibaba Cloud) come separately, in CNY, with stock per zone. USD per GPU-hour, list and market prices as published, not invoices. A provider that sells whole machines is divided by the machine's GPU count, so its figure includes the CPUs and memory of the machine. Spot, reserved and marketplace prices are different products from on-demand. A price does not say the GPU is in stock. Not purchasing advice.
| Argument | Type | Description |
|---|---|---|
| gpu | string | GPU id: b300, b200, h200, h100, a100-80gb, l40s, rtx-5090, rtx-4090, l20, a10, v100-32gb, v100, t4 |
| kind | string | Restrict to one kind of price |
{
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"properties": {
"gpu": {
"type": "string",
"maxLength": 40,
"description": "GPU id: b300, b200, h200, h100, a100-80gb, l40s, rtx-5090, rtx-4090, l20, a10, v100-32gb, v100, t4"
},
"kind": {
"enum": [
"ON_DEMAND",
"SPOT",
"RESERVED_1Y",
"RESERVED_3Y",
"MARKET"
],
"type": "string",
"description": "Restrict to one kind of price"
}
},
"additionalProperties": false
}Models with their official input/output list price, lowest channel and page. Optional vendor id or free-text filter. List prices as published by each channel at observed_at, USD per 1M tokens unless stated; China-site price lists are in CNY, listed separately and never converted. Standard (on-demand) plan unless service_plan says otherwise. The channel's own page is authoritative. Aggregator prices are the marketplace's own and never the reference. A missing price is unknown, not free. Not purchasing advice.
| Argument | Type | Description |
|---|---|---|
| q | string | Free text: model name, id or alias |
| vendor | string | Vendor id: anthropic, openai, google, meta, deepseek, alibaba, moonshot, zhipu, minimax |
{
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"properties": {
"q": {
"type": "string",
"maxLength": 80,
"description": "Free text: model name, id or alias"
},
"vendor": {
"type": "string",
"maxLength": 40,
"description": "Vendor id: anthropic, openai, google, meta, deepseek, alibaba, moonshot, zhipu, minimax"
}
},
"additionalProperties": false
}What hourly test requests measured per model and channel: time to first token, output speed, share of requests answered, and whether the charge matched the list price. Optional model and channel. Measured by AuraMatrix from one server in Los Angeles with one small request per model per hour; a sample, not a benchmark or a service-level figure. Requests our own account was refused are left out.
| Argument | Type | Description |
|---|---|---|
| model | string | Model id or alias; omit for all models, fastest first |
| channel | string | Channel id, e.g. openrouter, anthropic-api |
{
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"properties": {
"model": {
"type": "string",
"maxLength": 120,
"description": "Model id or alias; omit for all models, fastest first"
},
"channel": {
"type": "string",
"maxLength": 80,
"description": "Channel id, e.g. openrouter, anthropic-api"
}
},
"additionalProperties": false
}Observed changes between successive records of the same model and channel, plus vendor-announced changes with their effective dates. List prices as published by each channel at observed_at, USD per 1M tokens unless stated; China-site price lists are in CNY, listed separately and never converted. Standard (on-demand) plan unless service_plan says otherwise. The channel's own page is authoritative. Aggregator prices are the marketplace's own and never the reference. A missing price is unknown, not free. Not purchasing advice.
| Argument | Type | Description |
|---|---|---|
| days | integer | Look-back window, default 30 |
{
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"properties": {
"days": {
"type": "integer",
"maximum": 365,
"minimum": 1,
"description": "Look-back window, default 30"
}
},
"additionalProperties": false
}The trial indices: token price indices for flagship, mid and small models (100 on the base date) with their recent moves and announced changes, and the GPU-hour reference price per GPU model. Use it for "are model prices going up or down" rather than one model's price.
No arguments.
{
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"properties": {},
"additionalProperties": false
}Adds a price the user pays or was quoted, anonymously, to the benchmark of what buyers actually pay (a median is published once five reports exist for an item). The only tool here that stores anything. Call it ONLY after the user has explicitly agreed to share this price; never on your own initiative. Nothing identifying is stored: no account, no address. Do not put names of people or the user's company in supplier.
| Argument | Type | Description |
|---|---|---|
| item* | string | |
| kind* | string | |
| term | string | gpu only; default ON_DEMAND |
| price* | number | token: input price in USD per 1M tokens; gpu: price per GPU-hour |
| contract | string | |
| currency | string | gpu only |
| supplier | string | The provider or channel the price is from, if the user said |
| user_agreed* | boolean | true only if the user explicitly agreed to share this price |
| output_price | number |
We read every tool's name, description and inputs the way the model will, looking for instructions hidden from you. A clean result means no warning signs, not a guarantee. Scanned Nov 17, 2026.
| reasoning_tokens | integer | Thinking / reasoning tokens, counted separately from output_tokens |
| cache_write_tokens | integer | Tokens written to a cache |
| cached_input_tokens | integer | Tokens read from a cache |
{
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"model",
"input_tokens",
"output_tokens"
],
"properties": {
"at": {
"type": "string",
"maxLength": 40,
"description": "Request time, ISO 8601, for channels that price by time of day"
},
"model": {
"type": "string",
"maxLength": 120,
"minLength": 1,
"description": "Model id such as openai/gpt-6.1-sol, or an alias the channels use (gpt-6.1-sol, claude-opus-5-5). Case-insensitive."
},
"channel": {
"type": "string",
"maxLength": 80,
"description": "Channel id; default is the official channel"
},
"input_tokens": {
"type": "integer",
"maximum": 100000000000,
"minimum": 0,
"description": "Uncached input tokens"
},
"service_plan": {
"enum": [
"STANDARD",
"BATCH",
"FLEX",
"PRIORITY",
"ULTRAFAST"
],
"type": "string",
"description": "Default STANDARD"
},
"output_tokens": {
"type": "integer",
"maximum": 100000000000,
"minimum": 0,
"description": "Visible output tokens, without thinking tokens"
},
"reasoning_tokens": {
"type": "integer",
"maximum": 100000000000,
"minimum": 0,
"description": "Thinking / reasoning tokens, counted separately from output_tokens"
},
"cache_write_tokens": {
"type": "integer",
"maximum": 100000000000,
"minimum": 0,
"description": "Tokens written to a cache"
},
"cached_input_tokens": {
"type": "integer",
"maximum": 100000000000,
"minimum": 0,
"description": "Tokens read from a cache"
}
},
"additionalProperties": false
}| token only |
{
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"kind",
"item",
"price",
"user_agreed"
],
"properties": {
"item": {
"type": "string",
"maxLength": 120,
"minLength": 1
},
"kind": {
"enum": [
"token",
"gpu"
],
"type": "string"
},
"term": {
"enum": [
"ON_DEMAND",
"SPOT",
"MONTH_1",
"YEAR_1",
"YEAR_3"
],
"type": "string",
"description": "gpu only; default ON_DEMAND"
},
"price": {
"type": "number",
"maximum": 1000000,
"description": "token: input price in USD per 1M tokens; gpu: price per GPU-hour",
"exclusiveMinimum": 0
},
"contract": {
"enum": [
"quote",
"paying",
"committed",
"reseller"
],
"type": "string"
},
"currency": {
"enum": [
"USD",
"CNY"
],
"type": "string",
"description": "gpu only"
},
"supplier": {
"type": "string",
"maxLength": 80,
"description": "The provider or channel the price is from, if the user said"
},
"user_agreed": {
"type": "boolean",
"description": "true only if the user explicitly agreed to share this price"
},
"output_price": {
"type": "number",
"maximum": 1000000,
"description": "token only",
"exclusiveMinimum": 0
}
},
"additionalProperties": false
}