ModelCostComparison

io.github.Lazige由 GitHub 帳號 Lazige 發布v1.0.1更新於 Oct 26, 2026

Find AI model pricing, estimate token costs and compare offers. No API key required.

發布方自己的服務：model.tools — Zero-Auth Agent Utilities

已驗證Streamable HTTP可網頁執行金融 AI 與機器學習

服務詳細資訊

識別碼: io.github.Lazige/modelcostcomparison
傳輸: Streamable HTTP
執行環境: Web executable
信任: Verified
發布於: Sep 25, 2026

安全提示

MCP 服務會接收對話中的工具呼叫請求，並可能執行外部操作。啟用前請先了解該服務能做什麼，並只授予它所需的權限。

註冊中心已驗證此條目所屬的 GitHub 帳號 Lazige。這並不表明該帳號就是產品方。

SourceWeft 的檢查只涵蓋條目是否完整、端點是否回應；不審查伺服器的程式碼，也不審查其工具的行為。

在 README 中新增徽章

向讀者展示該 MCP 服務已收錄於 SourceWeft，徽章連結到本頁。

[![Listed on SourceWeft](https://sourceweft.com/badge/mcp/io.github.Lazige%2Fmodelcostcomparison.svg)](https://sourceweft.com/mcp/io.github.Lazige%2Fmodelcostcomparison?utm_source=readme&utm_medium=badge)

概覽

AI 產生的概覽

遠端 MCP 伺服器，用來查詢 AI 模型價格、估算美元文字 token 成本，並比較不同供應商的報價。

可以做什麼: 查詢供應商與模型，含報價數量及目前輸入／輸出價格區間
查看特定報價的費率、來源與審核日期
依單一報價計算輸入、輸出與快取輸入 token 的小計
在同一 token 用量與價格快照下比較 2–20 個報價
依指定用量對符合條件的已審核報價進行成本排序
適用情境: 適合在選型前讓助理估算或比較大模型 API 成本，或在預算討論中引用價格來源與審核日期。它是價格查詢與估算工具，不是帳單，也不評價模型品質。
執行需求: 需要一個遠端 Streamable HTTP 端點；不需 API 金鑰、本機套件或供應商憑證。服務有速率限制，遇到 HTTP 429 時應依 Retry-After 等待後重試。

安裝前請注意

結果僅為美元文字 token 小計估算，不含快取寫入、儲存、搜尋、工具費用與稅金，過期或未審核的報價無法計算。價格來自網站審核過的快照，而非即時抓取供應商資料。請求會傳送給第三方託管服務。

安裝

在 SourceWeft 中

開啟儀表板中的 ModelCostComparison，將其新增到工作區。
為需要使用其工具的對話啟用該服務。

Web executable，透過 Streamable HTTP。遠端服務在工作區中設定後即可從網頁執行環境執行。

交給代理程式安裝

把這段提示詞貼上到 Claude Code、Codex、Cursor 等代理程式，或 SourceWeft 對話中。代理程式會讀取該伺服器的安裝指南，先向你展示發布者、端點和登入方式，經你同意後再把它新增到你的客戶端。金鑰不會寫進任何命令。

閱讀 https://sourceweft.com/mcp/io.github.Lazige%2Fmodelcostcomparison/install.md，並按其中說明把這個 MCP 伺服器接入我的客戶端。修改任何設定之前，先告訴我它的發布者、端點和登入方式，等我確認。不要把 API 金鑰寫進命令或我會提交的檔案；問我金鑰應從哪裡取得。

檢視代理程式所依據的安裝指南

其他 MCP 客戶端

把它新增到你客戶端的 mcpServers 設定中。

{
  "mcpServers": {
    "modelcostcomparison": {
      "type": "http",
      "url": "https://modelcostcomparison.com/api/v1/mcp"
    }
  }
}

README

ModelCostComparison MCP

Measured by MCPmetrics:

[MCPmetrics: measured MCP protocol support]

Connect your AI client to the same pricing source and calculation engine used by ModelCostComparison.

Install in Cursor

Connect

Remote endpoint: https://modelcostcomparison.com/api/v1/mcp
Transport: Streamable HTTP. No API key or local package required.

Cursor

Use the install link above and confirm the server in Cursor. Alternatively, merge the contents of mcp.json into your Cursor MCP configuration, preserving existing servers. Enable ModelCostComparison in the agent's tools.

Claude

Open Connectors, choose Add custom connector, name it ModelCostComparison and paste the endpoint above. Connect it and enable its tools for your conversation. Organization settings may require an administrator to add it.

Other clients

Add a remote MCP server with the endpoint above and Streamable HTTP transport.

Tools

Tool	Purpose
`list_models`	Find model and exact tariff identifiers.
`get_model_pricing`	Get rates, sources and review dates.
`calculate_cost`	Calculate input, output and cached-input token costs.
`compare_model_costs`	Compare offers for the same token workload.
`find_cheapest_models`	Rank eligible reviewed offers for a token workload.
`list_providers`	Summarize providers, offers and current price ranges.

Try it

“Use ModelCostComparison to list available OpenAI models and show pricing sources and review dates.”
“Find reviewed offers for OpenAI, Anthropic and Google. Compare 1 million input tokens and 200,000 output tokens, with no cache and at most 8,000 input tokens per request.”
“For the selected offer, estimate 1 million input tokens including 500,000 cached input tokens, plus 100,000 output tokens. Maximum input per request is 8,000 tokens.”

Start with model discovery; use the exact offer IDs returned by the service. Missing or ambiguous workload details should be clarified before calculating.

Scope

Results are estimates of text-token subtotals in USD, not invoices or model-quality rankings. Cache writes, storage, tools, search and taxes are excluded. Expired or unreviewed offers cannot be calculated. Unsupported offers are reported separately in comparisons. Pricing is served from the website's reviewed deployment snapshot, not fetched from providers on every request.

The service is rate limited. On HTTP 429, wait for Retry-After before retrying. HTTP 503 indicates temporary unavailability. No provider credentials are needed, and tools do not call provider models.

This repository contains connection settings and documentation only. Pricing and calculations remain on the hosted service.

References

Official MCP Registry

Published as io.github.Lazige/modelcostcomparison, version 1.0.1. Registry entry. The registry metadata lives in server.json.

來源：README.md，提交 480bd4d

工具

list_models

Use first when you know a model name or provider but need exact pricing-offer IDs for pricing or calculations. Search recorded labels and API IDs; case, spacing and punctuation variants are accepted. Multiple tariffs are returned as separate choices. Defaults to current reviewed offers; check costing_supported before calculating. This discovers pricing offers, not model capability recommendations.

輸入參數limitquerycursorstatusprovider_idpricing_tier

get_model_pricing

Use after list_models when you need the rates, tariff identity, sources and freshness of one exact offer, without calculating a workload. Supply the returned model_id. Prices are recorded USD text-token rates; expired or unreviewed rates are informational only. Use calculate_cost for a workload estimate.

輸入參數model_id必填

calculate_cost

Use when you have one exact offer ID and token usage and want its USD text-token subtotal. Discover the ID with list_models first. Total input includes cached input; do not add cached tokens twice. Excludes cache writes, storage, search, tool fees and taxes. Requires current reviewed, supported rates; does not invoke a model. Use compare_model_costs for several known offers.

輸入參數usage必填model_id必填expected_dataset_revision

compare_model_costs

Use when you already have 2–20 exact offer IDs and want their USD text-token subtotals for the same usage and pricing snapshot. Equal token counts are a comparison assumption, not equivalent tokenization of the same text. Returns supported results and per-offer exclusions. Excludes cache writes, storage, search, tool fees and taxes; does not rank quality. Use find_cheapest_models to search without a shortlist.

輸入參數usage必填model_ids必填expected_dataset_revision

find_cheapest_models

Use when you know your token workload but have not selected offers: rank eligible current reviewed offers by USD text-token subtotal and return their exact IDs. Defaults to standard tariffs. Cheapest means within the eligible catalogue and supplied filters, not best quality or the entire market. Equal token counts do not imply tokenizer equivalence. Excludes cache writes, storage, search, tool fees and taxes; reports exclusion counts. Optional capacity filtering uses reviewed pricing input limits, not full context windows.

輸入參數limitusage必填provider_idpricing_tierexpected_dataset_revisionmin_input_tokens_per_request

list_providers

Use to discover provider IDs and catalogue coverage before filtering model searches. Returns distinct API model counts, offer counts and current supported input/output rate ranges in USD per million tokens. The two rate minima may belong to different offers. For a workload estimate use find_cheapest_models or calculate_cost; ranges are not capability rankings.

輸入參數provider_idpricing_tier