formatting

.changeset/bright-horses-double.md

@@ -1,5 +0,0 @@
----
-"claude-dev": patch
----
-
-Add correct cost and tokens info to Native OpenAI and DeepSeek providers

.changeset/calm-laws-camp.md

@@ -1,5 +0,0 @@
----
-"claude-dev": patch
----
-
-Update Claude Sonnet 35. -> 3.7 in README(s)

.changeset/dull-planets-battle.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": minor
+---
+
+Added timeout configuration for individual MCP servers.

.changeset/mighty-hounds-watch.md

@@ -1,5 +0,0 @@
----
-"claude-dev": patch
----
-
-Change how Cline finds the path to the user's Documents folder by querying xdg-user-dir on Linux systems.

README.md

@@ -32,7 +32,7 @@ English | <a href="https://github.com/cline/cline/blob/main/locales/es/README.md
 
 Meet Cline, an AI assistant that can use your **CLI** a**N**d **E**ditor.
 
-Thanks to [Claude 3.7 Sonnet's agentic coding capabilities](https://www.anthropic.com/claude/sonnet), Cline can handle complex software development tasks step-by-step. With tools that let him create & edit files, explore large projects, use the browser, and execute terminal commands (after you grant permission), he can assist you in ways that go beyond code completion or tech support. Cline can even use the Model Context Protocol (MCP) to create new tools and extend his own capabilities. While autonomous AI scripts traditionally run in sandboxed environments, this extension provides a human-in-the-loop GUI to approve every file change and terminal command, providing a safe and accessible way to explore the potential of agentic AI.
+Thanks to [Claude 3.5 Sonnet's agentic coding capabilities](https://www-cdn.anthropic.com/fed9cc193a14b84131812372d8d5857f8f304c52/Model_Card_Claude_3_Addendum.pdf), Cline can handle complex software development tasks step-by-step. With tools that let him create & edit files, explore large projects, use the browser, and execute terminal commands (after you grant permission), he can assist you in ways that go beyond code completion or tech support. Cline can even use the Model Context Protocol (MCP) to create new tools and extend his own capabilities. While autonomous AI scripts traditionally run in sandboxed environments, this extension provides a human-in-the-loop GUI to approve every file change and terminal command, providing a safe and accessible way to explore the potential of agentic AI.
 
 1. Enter your task and add images to convert mockups into functional apps or fix bugs with screenshots.
 2. Cline starts by analyzing your file structure & source code ASTs, running regex searches, and reading relevant files to get up to speed in existing projects. By carefully managing what information is added to context, Cline can provide valuable assistance even for large, complex projects without overwhelming the context window.

docs/PRIVACY.md

@@ -1,3 +1,3 @@
-View our Privacy Policy on our website.
+View our provacy policy on our website.
 
-[Privacy Policy](https://cline.bot/privacy)
+(Privacy Policy)[https://cline.bot/privacy]

docs/TERMS_OF_SERVICE.md

@@ -1,3 +1,3 @@
 View our Terms of Service on our website.
 
-[Terms of Service](https://cline.bot/tos)
+(Terms of Service)[https://cline.bot/tos]

locales/ar-sa/README.md

@@ -32,7 +32,7 @@
 
 التقى Cline، مساعد الذكاء الاصطناعي الذي يمكنه استخدام **سطر الأوامر** و **محرر النصوص** الخاص بك.
 
-بفضل [قدرات Claude 3.7 Sonnet على التعليمات البرمجية الوكيلة](https://www.anthropic.com/claude/sonnet)، يمكن لـ Cline التعامل مع مهام تطوير البرامج المعقدة خطوة بخطوة. مع الأدوات التي تسمح له بإنشاء وتعديل الملفات، واستكشاف المشاريع الكبيرة، واستخدام المتصفح، وتنفيذ أوامر الطرفية (بعد منحك الإذن)، يمكنه مساعدتك بطرق تتجاوز إكمال الكود أو الدعم الفني. يمكن لـ Cline أيضًا استخدام بروتوكول سياق النموذج (MCP) لإنشاء أدوات جديدة وتوسيع قدراته الخاصة. في حين تعمل النصوص البرمجية الآلية المستقلة تقليديًا في بيئات محاصرة، توفر هذه الإضافة واجهة رسومية لموافقة المستخدم على كل تغيير في الملف وأمر طرفية، مما يوفر طريقة آمنة وسهلة الاستخدام لاستكشاف إمكانات الذكاء الاصطناعي الوكيل.
+بفضل [قدرات Claude 3.5 Sonnet على التعليمات البرمجية الوكيلة](https://www-cdn.anthropic.com/fed9cc193a14b84131812372d8d5857f8f304c52/Model_Card_Claude_3_Addendum.pdf)، يمكن لـ Cline التعامل مع مهام تطوير البرامج المعقدة خطوة بخطوة. مع الأدوات التي تسمح له بإنشاء وتعديل الملفات، واستكشاف المشاريع الكبيرة، واستخدام المتصفح، وتنفيذ أوامر الطرفية (بعد منحك الإذن)، يمكنه مساعدتك بطرق تتجاوز إكمال الكود أو الدعم الفني. يمكن لـ Cline أيضًا استخدام بروتوكول سياق النموذج (MCP) لإنشاء أدوات جديدة وتوسيع قدراته الخاصة. في حين تعمل النصوص البرمجية الآلية المستقلة تقليديًا في بيئات محاصرة، توفر هذه الإضافة واجهة رسومية لموافقة المستخدم على كل تغيير في الملف وأمر طرفية، مما يوفر طريقة آمنة وسهلة الاستخدام لاستكشاف إمكانات الذكاء الاصطناعي الوكيل.
 
 1. أدخل مهمتك وأضف الصور لتحويل المحاكاة إلى تطبيقات وظيفية أو إصلاح الأخطاء مع لقطات الشاشة.
 2. يبدأ Cline بتحليل هيكل الملفات الخاصة بك وشجرة التعريف المصدرية، وإجراء عمليات بحث regex، وقراءة الملفات ذات الصلة للاطلاع على المشاريع الحالية. من خلال إدارة المعلومات التي يتم إضافتها إلى السياق بعناية، يمكن لـ Cline تقديم مساعدة قيمة حتى للمشاريع الكبيرة والمعقدة دون إرهاق نافذة السياق.

locales/de/README.md

@@ -28,7 +28,7 @@
 
 Lernen Sie Cline kennen, einen KI-Assistenten, der Ihre **CLI** u**N**d **E**ditor nutzen kann.
 
-Dank der [agentischen Codierungsfähigkeiten von Claude 3.7 Sonnet](https://www.anthropic.com/claude/sonnet) kann Cline komplexe Softwareentwicklungsaufgaben Schritt für Schritt bewältigen. Mit Werkzeugen, die ihm das Erstellen und Bearbeiten von Dateien, das Erkunden großer Projekte, die Nutzung des Browsers und das Ausführen von Terminalbefehlen (nach Ihrer Genehmigung) ermöglichen, kann er Ihnen auf eine Weise helfen, die über die Codevervollständigung oder technischen Support hinausgeht. Cline kann sogar das Model Context Protocol (MCP) verwenden, um neue Werkzeuge zu erstellen und seine eigenen Fähigkeiten zu erweitern. Während autonome KI-Skripte traditionell in sandboxed Umgebungen laufen, bietet diese Erweiterung eine Mensch-in-der-Schleife-GUI, um jede Dateiänderung und jeden Terminalbefehl zu genehmigen, was eine sichere und zugängliche Möglichkeit bietet, das Potenzial agentischer KI zu erkunden.
+Dank der [agentischen Codierungsfähigkeiten von Claude 3.5 Sonnet](https://www-cdn.anthropic.com/fed9cc193a14b84131812372d8d5857f8f304c52/Model_Card_Claude_3_Addendum.pdf) kann Cline komplexe Softwareentwicklungsaufgaben Schritt für Schritt bewältigen. Mit Werkzeugen, die ihm das Erstellen und Bearbeiten von Dateien, das Erkunden großer Projekte, die Nutzung des Browsers und das Ausführen von Terminalbefehlen (nach Ihrer Genehmigung) ermöglichen, kann er Ihnen auf eine Weise helfen, die über die Codevervollständigung oder technischen Support hinausgeht. Cline kann sogar das Model Context Protocol (MCP) verwenden, um neue Werkzeuge zu erstellen und seine eigenen Fähigkeiten zu erweitern. Während autonome KI-Skripte traditionell in sandboxed Umgebungen laufen, bietet diese Erweiterung eine Mensch-in-der-Schleife-GUI, um jede Dateiänderung und jeden Terminalbefehl zu genehmigen, was eine sichere und zugängliche Möglichkeit bietet, das Potenzial agentischer KI zu erkunden.
 
 1. Geben Sie Ihre Aufgabe ein und fügen Sie Bilder hinzu, um Mockups in funktionale Apps zu konvertieren oder Fehler mit Screenshots zu beheben.
 2. Cline beginnt mit der Analyse Ihrer Dateistruktur und Quellcode-ASTs, führt Regex-Suchen durch und liest relevante Dateien, um sich in bestehenden Projekten zurechtzufinden. Durch sorgfältiges Management der hinzugefügten Informationen kann Cline wertvolle Unterstützung auch bei großen, komplexen Projekten bieten, ohne das Kontextfenster zu überladen.

locales/es/README.md

@@ -28,7 +28,7 @@
 
 Conozca a Cline, un asistente de IA que puede usar su **CLI** y **E**ditor.
 
-Gracias a las [habilidades de codificación agencial de Claude 3.7 Sonnet](https://www.anthropic.com/claude/sonnet), Cline puede abordar tareas complejas de desarrollo de software paso a paso. Con herramientas que le permiten crear y editar archivos, explorar grandes proyectos, usar el navegador y ejecutar comandos de terminal (con su aprobación), puede ayudarle de una manera que va más allá de la autocompletación de código o el soporte técnico. Cline incluso puede usar el Model Context Protocol (MCP) para crear nuevas herramientas y expandir sus propias capacidades. Mientras que los scripts de IA autónomos tradicionalmente se ejecutan en entornos aislados, esta extensión ofrece una GUI con un humano en el bucle para aprobar cada cambio de archivo y comando de terminal, proporcionando una forma segura y accesible de explorar el potencial de la IA agencial.
+Gracias a las [habilidades de codificación agencial de Claude 3.5 Sonnet](https://www-cdn.anthropic.com/fed9cc193a14b84131812372d8d5857f8f304c52/Model_Card_Claude_3_Addendum.pdf), Cline puede abordar tareas complejas de desarrollo de software paso a paso. Con herramientas que le permiten crear y editar archivos, explorar grandes proyectos, usar el navegador y ejecutar comandos de terminal (con su aprobación), puede ayudarle de una manera que va más allá de la autocompletación de código o el soporte técnico. Cline incluso puede usar el Model Context Protocol (MCP) para crear nuevas herramientas y expandir sus propias capacidades. Mientras que los scripts de IA autónomos tradicionalmente se ejecutan en entornos aislados, esta extensión ofrece una GUI con un humano en el bucle para aprobar cada cambio de archivo y comando de terminal, proporcionando una forma segura y accesible de explorar el potencial de la IA agencial.
 
 1. Ingrese su tarea y agregue imágenes para convertir maquetas en aplicaciones funcionales o solucionar errores con capturas de pantalla.
 2. Cline comenzará analizando su estructura de archivos y ASTs de código fuente, realizando búsquedas Regex y leyendo archivos relevantes para orientarse en proyectos existentes. Al gestionar cuidadosamente la información agregada, Cline puede proporcionar asistencia valiosa incluso en proyectos grandes y complejos sin sobrecargar la ventana de contexto.

locales/ja/README.md

@@ -28,7 +28,7 @@
 
 Clineは、**CLI**と**エディター**を使用できるAIアシスタントです。
 
-[Claude 3.7 Sonnetのエージェント的コーディング機能](https://www.anthropic.com/claude/sonnet)のおかげで、Clineは複雑なソフトウェア開発タスクをステップバイステップで処理できます。ファイルの作成と編集、大規模プロジェクトの探索、ブラウザの使用、ターミナルコマンドの実行（許可後）などのツールを使用して、コード補完や技術サポートを超えた支援を提供します。Clineは、Model Context Protocol (MCP)を使用して新しいツールを作成し、自身の機能を拡張することもできます。自律的なAIスクリプトは通常サンドボックス環境で実行されますが、この拡張機能はファイル変更やターミナルコマンドを承認するための人間インターフェースを提供し、エージェント的AIの可能性を安全かつアクセスしやすい方法で探求できます。
+[Claude 3.5 Sonnetのエージェント的コーディング機能](https://www-cdn.anthropic.com/fed9cc193a14b84131812372d8d5857f8f304c52/Model_Card_Claude_3_Addendum.pdf)のおかげで、Clineは複雑なソフトウェア開発タスクをステップバイステップで処理できます。ファイルの作成と編集、大規模プロジェクトの探索、ブラウザの使用、ターミナルコマンドの実行（許可後）などのツールを使用して、コード補完や技術サポートを超えた支援を提供します。Clineは、Model Context Protocol (MCP)を使用して新しいツールを作成し、自身の機能を拡張することもできます。自律的なAIスクリプトは通常サンドボックス環境で実行されますが、この拡張機能はファイル変更やターミナルコマンドを承認するための人間インターフェースを提供し、エージェント的AIの可能性を安全かつアクセスしやすい方法で探求できます。
 
 1. タスクを入力し、モックアップを機能するアプリに変換したり、スクリーンショットでバグを修正したりします。
 2. Clineは、ファイル構造とソースコードASTの分析、正規表現検索の実行、関連ファイルの読み取りから始め、既存プロジェクトに精通します。コンテキストに追加される情報を慎重に管理することで、大規模で複雑なプロジェクトでもコンテキストウィンドウを圧倒することなく貴重な支援を提供できます。

locales/pt-BR/README.md

@@ -28,7 +28,7 @@
 
 Conheça o Cline: um assistente de IA que pode usar seu **CLI** e **Editor**.
 
-Graças às [habilidades avançadas do Claude 3.7 Sonnet](https://www.anthropic.com/claude/sonnet), o Cline pode lidar com tarefas complexas de desenvolvimento de software passo a passo. Com ferramentas que permitem criar e editar arquivos, explorar grandes projetos, usar o navegador e executar comandos no terminal (com sua aprovação), ele pode ajudar você de maneiras que vão além da inclusão de código ou suporte técnico. O Cline pode é capaz inclusive de usar o Model Context Protocol (MCP) para criar novas ferramentas e expandir seus próprios recursos. Embora os scripts de IA autônomas tradicionalmente sejam executados em ambientes isolados, esta extensão oferece uma GUI com um humano no circuito para aprovar cada alteração de arquivo e comando de terminal, fornecendo uma maneira segura e acessível de explorar todo o potencial da IA.
+Graças às [habilidades avançadas do Claude 3.5 Sonnet](https://www-cdn.anthropic.com/fed9cc193a14b84131812372d8d5857f8f304c52/Model_Card_Claude_3_Addendum.pdf), o Cline pode lidar com tarefas complexas de desenvolvimento de software passo a passo. Com ferramentas que permitem criar e editar arquivos, explorar grandes projetos, usar o navegador e executar comandos no terminal (com sua aprovação), ele pode ajudar você de maneiras que vão além da inclusão de código ou suporte técnico. O Cline pode é capaz inclusive de usar o Model Context Protocol (MCP) para criar novas ferramentas e expandir seus próprios recursos. Embora os scripts de IA autônomas tradicionalmente sejam executados em ambientes isolados, esta extensão oferece uma GUI com um humano no circuito para aprovar cada alteração de arquivo e comando de terminal, fornecendo uma maneira segura e acessível de explorar todo o potencial da IA.
 
 1. Insira sua tarefa e adicione imagens para transformar mockups em aplicativos funcionais ou corrigir erros através de capturas de tela.
 

locales/zh-cn/README.md

@@ -28,7 +28,7 @@
 
 认识 Cline，一个可以使用你的 **CLI** 和 **编辑器** 的 AI 助手。
 
-感谢 [Claude 3.7 Sonnet 的代理编码能力](https://www.anthropic.com/claude/sonnet)，Cline 可以一步步处理复杂的软件开发任务。通过允许他创建和编辑文件、探索大型项目、使用浏览器和执行终端命令（在你授予权限后），他可以提供超越代码完成或技术支持的帮助。Cline 甚至可以使用 Model Context Protocol (MCP) 创建新工具并扩展自己的能力。虽然自主 AI 脚本传统上在沙盒环境中运行，但此扩展提供了一个人机交互的 GUI 来批准每个文件更改和终端命令，提供了一种安全且可访问的方式来探索代理 AI 的潜力。
+感谢 [Claude 3.5 Sonnet 的代理编码能力](https://www-cdn.anthropic.com/fed9cc193a14b84131812372d8d5857f8f304c52/Model_Card_Claude_3_Addendum.pdf)，Cline 可以一步步处理复杂的软件开发任务。通过允许他创建和编辑文件、探索大型项目、使用浏览器和执行终端命令（在你授予权限后），他可以提供超越代码完成或技术支持的帮助。Cline 甚至可以使用 Model Context Protocol (MCP) 创建新工具并扩展自己的能力。虽然自主 AI 脚本传统上在沙盒环境中运行，但此扩展提供了一个人机交互的 GUI 来批准每个文件更改和终端命令，提供了一种安全且可访问的方式来探索代理 AI 的潜力。
 
 1. 输入你的任务并添加图像，将模型转换为功能应用程序或通过截图修复错误。
 2. Cline 首先分析你的文件结构和源代码 AST，运行正则表达式搜索，并阅读相关文件以了解现有项目。通过仔细管理添加到上下文中的信息，Cline 即使在大型复杂项目中也能提供有价值的帮助，而不会使上下文窗口过载。

locales/zh-tw/README.md

@@ -28,7 +28,7 @@
 
 認識 Cline，一個可以使用你的 **CLI** 和 **編輯器** 的 AI 助手。
 
-感謝 [Claude 3.7 Sonnet 的代理編碼能力](https://www.anthropic.com/claude/sonnet)，Cline 可以一步步處理複雜的軟件開發任務。通過允許他創建和編輯文件、探索大型項目、使用瀏覽器和執行終端命令（在你授予權限後），他可以提供超越代碼完成或技術支持的幫助。Cline 甚至可以使用 Model Context Protocol (MCP) 創建新工具並擴展自己的能力。雖然自主 AI 腳本傳統上在沙盒環境中運行，但此擴展提供了一個人機交互的 GUI 來批准每個文件更改和終端命令，提供了一種安全且可訪問的方式來探索代理 AI 的潛力。
+感謝 [Claude 3.5 Sonnet 的代理編碼能力](https://www-cdn.anthropic.com/fed9cc193a14b84131812372d8d5857f8f304c52/Model_Card_Claude_3_Addendum.pdf)，Cline 可以一步步處理複雜的軟件開發任務。通過允許他創建和編輯文件、探索大型項目、使用瀏覽器和執行終端命令（在你授予權限後），他可以提供超越代碼完成或技術支持的幫助。Cline 甚至可以使用 Model Context Protocol (MCP) 創建新工具並擴展自己的能力。雖然自主 AI 腳本傳統上在沙盒環境中運行，但此擴展提供了一個人機交互的 GUI 來批准每個文件更改和終端命令，提供了一種安全且可訪問的方式來探索代理 AI 的潛力。
 
 1. 輸入你的任務並添加圖像，將模型轉換為功能應用程序或通過截圖修復錯誤。
 2. Cline 首先分析你的文件結構和源代碼 AST，運行正則表達式搜索，並閱讀相關文件以了解現有項目。通過仔細管理添加到上下文中的信息，Cline 即使在大型複雜項目中也能提供有價值的幫助，而不會使上下文窗口過載。

src/api/providers/deepseek.ts

@@ -3,7 +3,6 @@ import OpenAI from "openai"
 import { withRetry } from "../retry"
 import { ApiHandler } from "../"
 import { ApiHandlerOptions, DeepSeekModelId, ModelInfo, deepSeekDefaultModelId, deepSeekModels } from "../../shared/api"
-import { calculateApiCostOpenAI } from "../../utils/cost"
 import { convertToOpenAiMessages } from "../transform/openai-format"
 import { ApiStream } from "../transform/stream"
 import { convertToR1Format } from "../transform/r1-format"
@@ -20,37 +19,6 @@ export class DeepSeekHandler implements ApiHandler {
 		})
 	}
 
-	private async *yieldUsage(info: ModelInfo, usage: OpenAI.Completions.CompletionUsage | undefined): ApiStream {
-		// Deepseek reports total input AND cache reads/writes,
-		// see context caching: https://api-docs.deepseek.com/guides/kv_cache)
-		// where the input tokens is the sum of the cache hits/misses, just like OpenAI.
-		// This affects:
-		// 1) context management truncation algorithm, and
-		// 2) cost calculation
-
-		// Deepseek usage includes extra fields.
-		// Safely cast the prompt token details section to the appropriate structure.
-		interface DeepSeekUsage extends OpenAI.CompletionUsage {
-			prompt_cache_hit_tokens?: number
-			prompt_cache_miss_tokens?: number
-		}
-		const deepUsage = usage as DeepSeekUsage
-
-		const inputTokens = deepUsage?.prompt_tokens || 0
-		const outputTokens = deepUsage?.completion_tokens || 0
-		const cacheReadTokens = deepUsage?.prompt_cache_hit_tokens || 0
-		const cacheWriteTokens = deepUsage?.prompt_cache_miss_tokens || 0
-		const totalCost = calculateApiCostOpenAI(info, inputTokens, outputTokens, cacheWriteTokens, cacheReadTokens)
-		yield {
-			type: "usage",
-			inputTokens: inputTokens,
-			outputTokens: outputTokens,
-			cacheWriteTokens: cacheWriteTokens,
-			cacheReadTokens: cacheReadTokens,
-			totalCost: totalCost,
-		}
-	}
-
 	@withRetry()
 	async *createMessage(systemPrompt: string, messages: Anthropic.Messages.MessageParam[]): ApiStream {
 		const model = this.getModel()
@@ -93,7 +61,15 @@ export class DeepSeekHandler implements ApiHandler {
 			}
 
 			if (chunk.usage) {
-				yield* this.yieldUsage(model.info, chunk.usage)
+				yield {
+					type: "usage",
+					inputTokens: chunk.usage.prompt_tokens || 0, // (deepseek reports total input AND cache reads/writes, see context caching: https://api-docs.deepseek.com/guides/kv_cache) where the input tokens is the sum of the cache hits/misses, while anthropic reports them as separate tokens. This is important to know for 1) context management truncation algorithm, and 2) cost calculation (NOTE: we report both input and cache stats but for now set input price to 0 since all the cost calculation will be done using cache hits/misses)
+					outputTokens: chunk.usage.completion_tokens || 0,
+					// @ts-ignore-next-line
+					cacheReadTokens: chunk.usage.prompt_cache_hit_tokens || 0,
+					// @ts-ignore-next-line
+					cacheWriteTokens: chunk.usage.prompt_cache_miss_tokens || 0,
+				}
 			}
 		}
 	}

src/api/providers/openai-native.ts

@@ -10,7 +10,6 @@ import {
 	openAiNativeModels,
 } from "../../shared/api"
 import { convertToOpenAiMessages } from "../transform/openai-format"
-import { calculateApiCostOpenAI } from "../../utils/cost"
 import { ApiStream } from "../transform/stream"
 import { ChatCompletionReasoningEffort } from "openai/resources/chat/completions.mjs"
 
@@ -25,47 +24,31 @@ export class OpenAiNativeHandler implements ApiHandler {
 		})
 	}
 
-	private async *yieldUsage(info: ModelInfo, usage: OpenAI.Completions.CompletionUsage | undefined): ApiStream {
-		const inputTokens = usage?.prompt_tokens || 0
-		const outputTokens = usage?.completion_tokens || 0
-		const cacheReadTokens = usage?.prompt_tokens_details?.cached_tokens || 0
-		const cacheWriteTokens = 0
-		const totalCost = calculateApiCostOpenAI(info, inputTokens, outputTokens, cacheWriteTokens, cacheReadTokens)
-		yield {
-			type: "usage",
-			inputTokens: inputTokens,
-			outputTokens: outputTokens,
-			cacheWriteTokens: cacheWriteTokens,
-			cacheReadTokens: cacheReadTokens,
-			totalCost: totalCost,
-		}
-	}
-
 	@withRetry()
 	async *createMessage(systemPrompt: string, messages: Anthropic.Messages.MessageParam[]): ApiStream {
-		const model = this.getModel()
-
-		switch (model.id) {
+		switch (this.getModel().id) {
 			case "o1":
 			case "o1-preview":
 			case "o1-mini": {
 				// o1 doesnt support streaming, non-1 temp, or system prompt
 				const response = await this.client.chat.completions.create({
-					model: model.id,
+					model: this.getModel().id,
 					messages: [{ role: "user", content: systemPrompt }, ...convertToOpenAiMessages(messages)],
 				})
 				yield {
 					type: "text",
 					text: response.choices[0]?.message.content || "",
 				}
-
-				yield* this.yieldUsage(model.info, response.usage)
-
+				yield {
+					type: "usage",
+					inputTokens: response.usage?.prompt_tokens || 0,
+					outputTokens: response.usage?.completion_tokens || 0,
+				}
 				break
 			}
 			case "o3-mini": {
 				const stream = await this.client.chat.completions.create({
-					model: model.id,
+					model: this.getModel().id,
 					messages: [{ role: "developer", content: systemPrompt }, ...convertToOpenAiMessages(messages)],
 					stream: true,
 					stream_options: { include_usage: true },
@@ -80,15 +63,18 @@ export class OpenAiNativeHandler implements ApiHandler {
 						}
 					}
 					if (chunk.usage) {
-						// Only last chunk contains usage
-						yield* this.yieldUsage(model.info, chunk.usage)
+						yield {
+							type: "usage",
+							inputTokens: chunk.usage.prompt_tokens || 0,
+							outputTokens: chunk.usage.completion_tokens || 0,
+						}
 					}
 				}
 				break
 			}
 			default: {
 				const stream = await this.client.chat.completions.create({
-					model: model.id,
+					model: this.getModel().id,
 					// max_completion_tokens: this.getModel().info.maxTokens,
 					temperature: 0,
 					messages: [{ role: "system", content: systemPrompt }, ...convertToOpenAiMessages(messages)],
@@ -104,9 +90,14 @@ export class OpenAiNativeHandler implements ApiHandler {
 							text: delta.content,
 						}
 					}
+
+					// contains a null value except for the last chunk which contains the token usage statistics for the entire request
 					if (chunk.usage) {
-						// Only last chunk contains usage
-						yield* this.yieldUsage(model.info, chunk.usage)
+						yield {
+							type: "usage",
+							inputTokens: chunk.usage.prompt_tokens || 0,
+							outputTokens: chunk.usage.completion_tokens || 0,
+						}
 					}
 				}
 			}

src/api/providers/requesty.ts

@@ -2,7 +2,13 @@ import { Anthropic } from "@anthropic-ai/sdk"
 import OpenAI from "openai"
 import { withRetry } from "../retry"
 import { calculateApiCostOpenAI } from "../../utils/cost"
-import { ApiHandlerOptions, ModelInfo, openAiModelInfoSaneDefaults, requestyDefaultModelId, requestyDefaultModelInfo  } from "../../shared/api"
+import {
+	ApiHandlerOptions,
+	ModelInfo,
+	openAiModelInfoSaneDefaults,
+	requestyDefaultModelId,
+	requestyDefaultModelInfo,
+} from "../../shared/api"
 import { ApiHandler } from "../index"
 import { convertToOpenAiMessages } from "../transform/openai-format"
 import { ApiStream } from "../transform/stream"
@@ -71,19 +77,13 @@ export class RequestyHandler implements ApiHandler {
 
 			if (chunk.usage) {
 				const usage = chunk.usage as RequestyUsage
-				const inputTokens = usage.prompt_tokens || 0
-				const outputTokens = usage.completion_tokens || 0
-				const cacheWriteTokens = usage.prompt_tokens_details?.caching_tokens || undefined
-				const cacheReadTokens = usage.prompt_tokens_details?.cached_tokens || undefined
-				const totalCost = 0 // TODO: Replace with calculateApiCostOpenAI(model.info, inputTokens, outputTokens, cacheWriteTokens, cacheReadTokens)
-
 				yield {
 					type: "usage",
-					inputTokens: inputTokens,
-					outputTokens: outputTokens,
-					cacheWriteTokens: cacheWriteTokens,
-					cacheReadTokens: cacheReadTokens,
-					totalCost: totalCost,
+					inputTokens: usage.prompt_tokens || 0,
+					outputTokens: usage.completion_tokens || 0,
+					cacheWriteTokens: usage.prompt_tokens_details?.caching_tokens || undefined,
+					cacheReadTokens: usage.prompt_tokens_details?.cached_tokens || undefined,
+					totalCost: usage.total_cost || undefined,
 				}
 			}
 		}

src/api/providers/vscode-lm.ts

@@ -1,7 +1,7 @@
 import { Anthropic } from "@anthropic-ai/sdk"
 import * as vscode from "vscode"
 import { ApiHandler, SingleCompletionHandler } from "../"
-import { calculateApiCostAnthropic } from "../../utils/cost"
+import { calculateApiCost } from "../../utils/cost"
 import { ApiStream } from "../transform/stream"
 import { convertToVsCodeLmMessages } from "../transform/vscode-lm-format"
 import { SELECTOR_SEPARATOR, stringifyVsCodeLmModelSelector } from "../../shared/vsCodeSelectorUtils"
@@ -525,7 +525,7 @@ export class VsCodeLmHandler implements ApiHandler, SingleCompletionHandler {
 				type: "usage",
 				inputTokens: totalInputTokens,
 				outputTokens: totalOutputTokens,
-				totalCost: calculateApiCostAnthropic(this.getModel().info, totalInputTokens, totalOutputTokens),
+				totalCost: calculateApiCost(this.getModel().info, totalInputTokens, totalOutputTokens),
 			}
 		} catch (error: unknown) {
 			this.ensureCleanState()

src/core/Cline.ts

@@ -47,7 +47,7 @@ import {
 import { getApiMetrics } from "../shared/getApiMetrics"
 import { HistoryItem } from "../shared/HistoryItem"
 import { ClineAskResponse, ClineCheckpointRestore } from "../shared/WebviewMessage"
-import { calculateApiCostAnthropic } from "../utils/cost"
+import { calculateApiCost } from "../utils/cost"
 import { fileExistsAtPath } from "../utils/fs"
 import { arePathsEqual, getReadablePath } from "../utils/path"
 import { fixModelHtmlEscaping, removeInvalidChars } from "../utils/string"
@@ -3113,13 +3113,7 @@ export class Cline {
 					cacheReads: cacheReadTokens,
 					cost:
 						totalCost ??
-						calculateApiCostAnthropic(
-							this.api.getModel().info,
-							inputTokens,
-							outputTokens,
-							cacheWriteTokens,
-							cacheReadTokens,
-						),
+						calculateApiCost(this.api.getModel().info, inputTokens, outputTokens, cacheWriteTokens, cacheReadTokens),
 					cancelReason,
 					streamingFailedMessage,
 				} satisfies ClineApiReqInfo)

src/core/prompts/system.ts

@@ -3,6 +3,7 @@ import os from "os"
 import osName from "os-name"
 import { McpHub } from "../../services/mcp/McpHub"
 import { BrowserSettings } from "../../shared/BrowserSettings"
+import { DEFAULT_MCP_TIMEOUT_SECONDS } from "../../shared/mcp"
 
 export const SYSTEM_PROMPT = async (
 	cwd: string,
@@ -759,8 +760,9 @@ IMPORTANT: Regardless of what else you see in the MCP settings file, you must de
       "args": ["/path/to/weather-server/build/index.js"],
       "env": {
         "OPENWEATHER_API_KEY": "user-provided-api-key"
-      }
-    },
+      },
+      "timeout": ${DEFAULT_MCP_TIMEOUT_SECONDS}, // ${DEFAULT_MCP_TIMEOUT_SECONDS} seconds should be set as the default timeout for all MCP servers
+    }
   }
 }
 \`\`\`

src/core/webview/ClineProvider.ts

@@ -542,7 +542,6 @@ export class ClineProvider implements vscode.WebviewViewProvider {
 							telemetryService.updateTelemetryState(isOptedIn)
 						})
 
-            
 						// post last cached models in case the call to endpoint fails
 						this.readDynamicProviderModels(GlobalFileNames.requestyModels).then((requestyModels) => {
 							if (requestyModels) {
@@ -552,7 +551,7 @@ export class ClineProvider implements vscode.WebviewViewProvider {
 								})
 							}
 						})
-            
+
 						// gui relies on model info to be up-to-date to provide the most accurate pricing, so we need to fetch the latest details on launch.
 						// we do this for all users since many users switch between api providers and if they were to switch back to openrouter it would be showing outdated model info if we hadn't retrieved the latest at this point
 						// (see normalizeApiConfiguration > openrouter)
@@ -569,7 +568,7 @@ export class ClineProvider implements vscode.WebviewViewProvider {
 								}
 							}
 						})
-            
+
 						break
 					case "newTask":
 						// Code that should run in response to the hello message command
@@ -974,6 +973,16 @@ export class ClineProvider implements vscode.WebviewViewProvider {
 						}
 						break
 					}
+					case "updateMcpTimeout": {
+						try {
+							if (message.serverName && message.timeout) {
+								await this.mcpHub?.updateServerTimeout(message.serverName, message.timeout)
+							}
+						} catch (error) {
+							console.error(`Failed to update timeout for server ${message.serverName}:`, error)
+						}
+						break
+					}
 					case "openExtensionSettings": {
 						const settingsFilter = message.text || ""
 						await vscode.commands.executeCommand(
@@ -1187,38 +1196,21 @@ export class ClineProvider implements vscode.WebviewViewProvider {
 
 	async getDocumentsPath(): Promise<string> {
 		if (process.platform === "win32") {
+			// If the user is running Win 7/Win Server 2008 r2+, we want to get the correct path to their Documents directory.
 			try {
 				const { stdout: docsPath } = await execa("powershell", [
 					"-NoProfile", // Ignore user's PowerShell profile(s)
 					"-Command",
 					"[System.Environment]::GetFolderPath([System.Environment+SpecialFolder]::MyDocuments)",
 				])
-				const trimmedPath = docsPath.trim()
-				if (trimmedPath) {
-					return trimmedPath
-				}
+				return docsPath.trim()
 			} catch (err) {
 				console.error("Failed to retrieve Windows Documents path. Falling back to homedir/Documents.")
+				return path.join(os.homedir(), "Documents")
 			}
-		} else if (process.platform === "linux") {
-			try {
-				// First check if xdg-user-dir exists
-				await execa("which", ["xdg-user-dir"])
-
-				// If it exists, try to get XDG documents path
-				const { stdout } = await execa("xdg-user-dir", ["DOCUMENTS"])
-				const trimmedPath = stdout.trim()
-				if (trimmedPath) {
-					return trimmedPath
-				}
-			} catch {
-				// Log error but continue to fallback
-				console.error("Failed to retrieve XDG Documents path. Falling back to homedir/Documents.")
-			}
+		} else {
+			return path.join(os.homedir(), "Documents") // On POSIX (macOS, Linux, etc.), assume ~/Documents by default (existing behavior, but may want to implement similar logic here)
 		}
-
-		// Default fallback for all platforms
-		return path.join(os.homedir(), "Documents")
 	}
 
 	async ensureMcpServersDirectoryExists(): Promise<string> {

src/services/mcp/McpHub.ts

@@ -16,6 +16,7 @@ import * as vscode from "vscode"
 import { z } from "zod"
 import { ClineProvider, GlobalFileNames } from "../../core/webview/ClineProvider"
 import {
+	DEFAULT_MCP_TIMEOUT_SECONDS,
 	McpMode,
 	McpResource,
 	McpResourceResponse,
@@ -23,10 +24,11 @@ import {
 	McpServer,
 	McpTool,
 	McpToolCallResponse,
+	MIN_MCP_TIMEOUT_SECONDS,
 } from "../../shared/mcp"
 import { fileExistsAtPath } from "../../utils/fs"
 import { arePathsEqual } from "../../utils/path"
-
+import { secondsToMs } from "../../utils/time"
 export type McpConnection = {
 	server: McpServer
 	client: Client
@@ -35,13 +37,13 @@ export type McpConnection = {
 
 const AutoApproveSchema = z.array(z.string()).default([])
 
-// StdioServerParameters
 const StdioConfigSchema = z.object({
 	command: z.string(),
 	args: z.array(z.string()).optional(),
 	env: z.record(z.string()).optional(),
 	autoApprove: AutoApproveSchema.optional(),
 	disabled: z.boolean().optional(),
+	timeout: z.number().min(MIN_MCP_TIMEOUT_SECONDS).optional().default(DEFAULT_MCP_TIMEOUT_SECONDS),
 })
 
 const McpSettingsSchema = z.object({
@@ -242,28 +244,6 @@ export class McpHub {
 			}
 			transport.start = async () => {} // No-op now, .connect() won't fail
 
-			// // Set up notification handlers
-			// client.setNotificationHandler(
-			// 	// @ts-ignore-next-line
-			// 	{ method: "notifications/tools/list_changed" },
-			// 	async () => {
-			// 		console.log(`Tools changed for server: ${name}`)
-			// 		connection.server.tools = await this.fetchTools(name)
-			// 		await this.notifyWebviewOfServerChanges()
-			// 	},
-			// )
-
-			// client.setNotificationHandler(
-			// 	// @ts-ignore-next-line
-			// 	{ method: "notifications/resources/list_changed" },
-			// 	async () => {
-			// 		console.log(`Resources changed for server: ${name}`)
-			// 		connection.server.resources = await this.fetchResources(name)
-			// 		connection.server.resourceTemplates = await this.fetchResourceTemplates(name)
-			// 		await this.notifyWebviewOfServerChanges()
-			// 	},
-			// )
-
 			// Connect
 			await client.connect(transport)
 			connection.server.status = "connected"
@@ -343,10 +323,6 @@ export class McpHub {
 		const connection = this.connections.find((conn) => conn.server.name === name)
 		if (connection) {
 			try {
-				// connection.client.removeNotificationHandler("notifications/tools/list_changed")
-				// connection.client.removeNotificationHandler("notifications/resources/list_changed")
-				// connection.client.removeNotificationHandler("notifications/stderr")
-				// connection.client.removeNotificationHandler("notifications/stderr")
 				await connection.transport.close()
 				await connection.client.close()
 			} catch (error) {
@@ -563,6 +539,7 @@ export class McpHub {
 		if (connection.server.disabled) {
 			throw new Error(`Server "${serverName}" is disabled`)
 		}
+
 		return await connection.client.request(
 			{
 				method: "resources/read",
@@ -586,6 +563,17 @@ export class McpHub {
 			throw new Error(`Server "${serverName}" is disabled and cannot be used`)
 		}
 
+		let timeout = secondsToMs(DEFAULT_MCP_TIMEOUT_SECONDS) // sdk expects ms
+
+		try {
+			const config = JSON.parse(connection.server.config)
+			const parsedConfig = StdioConfigSchema.parse(config)
+			timeout = secondsToMs(parsedConfig.timeout)
+		} catch (error) {
+			console.error(`Failed to parse timeout configuration for server ${serverName}: ${error}`)
+			// Continue with default timeout
+		}
+
 		return await connection.client.request(
 			{
 				method: "tools/call",
@@ -595,6 +583,9 @@ export class McpHub {
 				},
 			},
 			CallToolResultSchema,
+			{
+				timeout,
+			},
 		)
 	}
 
@@ -663,6 +654,47 @@ export class McpHub {
 		}
 	}
 
+	public async updateServerTimeout(serverName: string, timeout: number): Promise<void> {
+		try {
+			// Validate timeout against schema
+			const setConfigResult = StdioConfigSchema.shape.timeout.safeParse(timeout)
+			if (!setConfigResult.success) {
+				throw new Error(`Invalid timeout value: ${timeout}. Must be at minimum ${MIN_MCP_TIMEOUT_SECONDS} seconds.`)
+			}
+
+			const settingsPath = await this.getMcpSettingsFilePath()
+			const content = await fs.readFile(settingsPath, "utf-8")
+			const config = JSON.parse(content)
+
+			if (!config.mcpServers?.[serverName]) {
+				throw new Error(`Server "${serverName}" not found in settings`)
+			}
+
+			// Update the timeout in the config
+			config.mcpServers[serverName] = {
+				...config.mcpServers[serverName],
+				timeout,
+			}
+
+			// Write updated config back to file
+			await fs.writeFile(settingsPath, JSON.stringify(config, null, 2))
+
+			// Update server connections to apply the new timeout
+			await this.updateServerConnections(config.mcpServers)
+
+			vscode.window.showInformationMessage(`Updated timeout to ${timeout} seconds`)
+		} catch (error) {
+			console.error("Failed to update server timeout:", error)
+			if (error instanceof Error) {
+				console.error("Error details:", error.message, error.stack)
+			}
+			vscode.window.showErrorMessage(
+				`Failed to update server timeout: ${error instanceof Error ? error.message : String(error)}`,
+			)
+			throw error
+		}
+	}
+
 	async dispose(): Promise<void> {
 		this.removeAllFileWatchers()
 		for (const connection of this.connections) {

src/shared/WebviewMessage.ts

@@ -70,6 +70,7 @@ export interface WebviewMessage {
 	chatSettings?: ChatSettings
 	chatContent?: ChatContent
 	mcpId?: string
+	timeout?: number // For updateMcpTimeout
 
 	// For toggleToolAutoApprove
 	serverName?: string

src/shared/api.ts

@@ -411,10 +411,9 @@ export const openAiNativeModels = {
 		maxTokens: 100_000,
 		contextWindow: 200_000,
 		supportsImages: false,
-		supportsPromptCache: true,
+		supportsPromptCache: false,
 		inputPrice: 1.1,
 		outputPrice: 4.4,
-		cacheReadsPrice: 0.55,
 	},
 	// don't support tool use yet
 	o1: {
@@ -424,43 +423,38 @@ export const openAiNativeModels = {
 		supportsPromptCache: false,
 		inputPrice: 15,
 		outputPrice: 60,
-		cacheReadsPrice: 7.5,
 	},
 	"o1-preview": {
 		maxTokens: 32_768,
 		contextWindow: 128_000,
 		supportsImages: true,
-		supportsPromptCache: true,
+		supportsPromptCache: false,
 		inputPrice: 15,
 		outputPrice: 60,
-		cacheReadsPrice: 7.5,
 	},
 	"o1-mini": {
 		maxTokens: 65_536,
 		contextWindow: 128_000,
 		supportsImages: true,
-		supportsPromptCache: true,
+		supportsPromptCache: false,
 		inputPrice: 1.1,
 		outputPrice: 4.4,
-		cacheReadsPrice: 0.55,
 	},
 	"gpt-4o": {
 		maxTokens: 4_096,
 		contextWindow: 128_000,
 		supportsImages: true,
-		supportsPromptCache: true,
+		supportsPromptCache: false,
 		inputPrice: 2.5,
 		outputPrice: 10,
-		cacheReadsPrice: 1.25,
 	},
 	"gpt-4o-mini": {
 		maxTokens: 16_384,
 		contextWindow: 128_000,
 		supportsImages: true,
-		supportsPromptCache: true,
+		supportsPromptCache: false,
 		inputPrice: 0.15,
 		outputPrice: 0.6,
-		cacheReadsPrice: 0.075,
 	},
 	"gpt-4.5-preview": {
 		maxTokens: 16_384,
@@ -486,8 +480,8 @@ export const deepSeekModels = {
 		maxTokens: 8_000,
 		contextWindow: 64_000,
 		supportsImages: false,
-		supportsPromptCache: true,
-		inputPrice: 0.27,
+		supportsPromptCache: true, // supports context caching, but not in the way anthropic does it (deepseek reports input tokens and reads/writes in the same usage report) FIXME: we need to show users cache stats how deepseek does it
+		inputPrice: 0, // technically there is no input price, it's all either a cache hit or miss (ApiOptions will not show this)
 		outputPrice: 1.1,
 		cacheWritesPrice: 0.27,
 		cacheReadsPrice: 0.07,
@@ -496,8 +490,8 @@ export const deepSeekModels = {
 		maxTokens: 8_000,
 		contextWindow: 64_000,
 		supportsImages: false,
-		supportsPromptCache: true,
-		inputPrice: 0.55,
+		supportsPromptCache: true, // supports context caching, but not in the way anthropic does it (deepseek reports input tokens and reads/writes in the same usage report) FIXME: we need to show users cache stats how deepseek does it
+		inputPrice: 0, // technically there is no input price, it's all either a cache hit or miss (ApiOptions will not show this)
 		outputPrice: 2.19,
 		cacheWritesPrice: 0.55,
 		cacheReadsPrice: 0.14,
@@ -897,4 +891,3 @@ export const xaiModels = {
 		description: "X AI's Grok Beta model (legacy) with 131K context window",
 	},
 } as const satisfies Record<string, ModelInfo>
-

src/shared/mcp.ts

@@ -1,3 +1,6 @@
+export const DEFAULT_MCP_TIMEOUT_SECONDS = 60 // matches Anthropic's default timeout in their MCP SDK
+export const MIN_MCP_TIMEOUT_SECONDS = 1
+
 export type McpMode = "full" | "server-use-only" | "off"
 
 export type McpServer = {
@@ -9,6 +12,7 @@ export type McpServer = {
 	resources?: McpResource[]
 	resourceTemplates?: McpResourceTemplate[]
 	disabled?: boolean
+	timeout?: number
 }
 
 export type McpTool = {

src/utils/cost.test.ts

@@ -1,10 +1,10 @@
 import { describe, it } from "mocha"
 import "should"
-import { calculateApiCostAnthropic, calculateApiCostOpenAI } from "./cost"
+import { calculateApiCost } from "./cost"
 import { ModelInfo } from "../shared/api"
 
 describe("Cost Utilities", () => {
-	describe("calculateApiCostAnthropic", () => {
+	describe("calculateApiCost", () => {
 		it("should calculate basic input/output costs", () => {
 			const modelInfo: ModelInfo = {
 				supportsPromptCache: false,
@@ -12,7 +12,7 @@ describe("Cost Utilities", () => {
 				outputPrice: 15.0, // $15 per million tokens
 			}
 
-			const cost = calculateApiCostAnthropic(modelInfo, 1000, 500)
+			const cost = calculateApiCost(modelInfo, 1000, 500)
 			// Input: (3.0 / 1_000_000) * 1000 = 0.003
 			// Output: (15.0 / 1_000_000) * 500 = 0.0075
 			// Total: 0.003 + 0.0075 = 0.0105
@@ -25,7 +25,7 @@ describe("Cost Utilities", () => {
 				// No prices specified
 			}
 
-			const cost = calculateApiCostAnthropic(modelInfo, 1000, 500)
+			const cost = calculateApiCost(modelInfo, 1000, 500)
 			cost.should.equal(0)
 		})
 
@@ -42,7 +42,7 @@ describe("Cost Utilities", () => {
 				cacheReadsPrice: 0.3,
 			}
 
-			const cost = calculateApiCostAnthropic(modelInfo, 2000, 1000, 1500, 500)
+			const cost = calculateApiCost(modelInfo, 2000, 1000, 1500, 500)
 			// Cache writes: (3.75 / 1_000_000) * 1500 = 0.005625
 			// Cache reads: (0.3 / 1_000_000) * 500 = 0.00015
 			// Input: (3.0 / 1_000_000) * 2000 = 0.006
@@ -60,68 +60,7 @@ describe("Cost Utilities", () => {
 				cacheReadsPrice: 0.3,
 			}
 
-			const cost = calculateApiCostAnthropic(modelInfo, 0, 0, 0, 0)
-			cost.should.equal(0)
-		})
-	})
-
-	describe("calculateApiCostOpenAI", () => {
-		it("should calculate basic input/output costs", () => {
-			const modelInfo: ModelInfo = {
-				supportsPromptCache: false,
-				inputPrice: 3.0, // $3 per million tokens
-				outputPrice: 15.0, // $15 per million tokens
-			}
-
-			const cost = calculateApiCostOpenAI(modelInfo, 1000, 500)
-			// Input: (3.0 / 1_000_000) * 1000 = 0.003
-			// Output: (15.0 / 1_000_000) * 500 = 0.0075
-			// Total: 0.003 + 0.0075 = 0.0105
-			cost.should.equal(0.0105)
-		})
-
-		it("should handle missing prices", () => {
-			const modelInfo: ModelInfo = {
-				supportsPromptCache: true,
-				// No prices specified
-			}
-
-			const cost = calculateApiCostOpenAI(modelInfo, 1000, 500)
-			cost.should.equal(0)
-		})
-
-		it("should use real model configuration (Claude 3.5 Sonnet)", () => {
-			const modelInfo: ModelInfo = {
-				maxTokens: 8192,
-				contextWindow: 200_000,
-				supportsImages: true,
-				supportsComputerUse: true,
-				supportsPromptCache: true,
-				inputPrice: 3.0,
-				outputPrice: 15.0,
-				cacheWritesPrice: 3.75,
-				cacheReadsPrice: 0.3,
-			}
-
-			const cost = calculateApiCostOpenAI(modelInfo, 2100, 1000, 1500, 500)
-			// Cache writes: (3.75 / 1_000_000) * 1500 = 0.005625
-			// Cache reads: (0.3 / 1_000_000) * 500 = 0.00015
-			// Input: (3.0 / 1_000_000) * (2100 - 1500 - 500) = 0.0003
-			// Output: (15.0 / 1_000_000) * 1000 = 0.015
-			// Total: 0.005625 + 0.00015 + 0.0003 + 0.015 = 0.021075
-			cost.should.equal(0.021075)
-		})
-
-		it("should handle zero token counts", () => {
-			const modelInfo: ModelInfo = {
-				supportsPromptCache: true,
-				inputPrice: 3.0,
-				outputPrice: 15.0,
-				cacheWritesPrice: 3.75,
-				cacheReadsPrice: 0.3,
-			}
-
-			const cost = calculateApiCostOpenAI(modelInfo, 0, 0, 0, 0)
+			const cost = calculateApiCost(modelInfo, 0, 0, 0, 0)
 			cost.should.equal(0)
 		})
 	})

src/utils/cost.ts

@@ -1,49 +1,24 @@
 import { ModelInfo } from "../shared/api"
 
-function calculateApiCostInternal(
+export function calculateApiCost(
 	modelInfo: ModelInfo,
 	inputTokens: number,
 	outputTokens: number,
-	cacheCreationInputTokens: number,
-	cacheReadInputTokens: number,
+	cacheCreationInputTokens?: number,
+	cacheReadInputTokens?: number,
 ): number {
-	const cacheWritesCost = ((modelInfo.cacheWritesPrice || 0) / 1_000_000) * cacheCreationInputTokens
-	const cacheReadsCost = ((modelInfo.cacheReadsPrice || 0) / 1_000_000) * cacheReadInputTokens
+	const modelCacheWritesPrice = modelInfo.cacheWritesPrice
+	let cacheWritesCost = 0
+	if (cacheCreationInputTokens && modelCacheWritesPrice) {
+		cacheWritesCost = (modelCacheWritesPrice / 1_000_000) * cacheCreationInputTokens
+	}
+	const modelCacheReadsPrice = modelInfo.cacheReadsPrice
+	let cacheReadsCost = 0
+	if (cacheReadInputTokens && modelCacheReadsPrice) {
+		cacheReadsCost = (modelCacheReadsPrice / 1_000_000) * cacheReadInputTokens
+	}
 	const baseInputCost = ((modelInfo.inputPrice || 0) / 1_000_000) * inputTokens
 	const outputCost = ((modelInfo.outputPrice || 0) / 1_000_000) * outputTokens
 	const totalCost = cacheWritesCost + cacheReadsCost + baseInputCost + outputCost
 	return totalCost
 }
-
-// For Anthropic compliant usage, the input tokens count does NOT include the cached tokens
-export function calculateApiCostAnthropic(
-	modelInfo: ModelInfo,
-	inputTokens: number,
-	outputTokens: number,
-	cacheCreationInputTokens?: number,
-	cacheReadInputTokens?: number,
-): number {
-	const cacheCreationInputTokensNum = cacheCreationInputTokens || 0
-	const cacheReadInputTokensNum = cacheReadInputTokens || 0
-	return calculateApiCostInternal(modelInfo, inputTokens, outputTokens, cacheCreationInputTokensNum, cacheReadInputTokensNum)
-}
-
-// For OpenAI compliant usage, the input tokens count INCLUDES the cached tokens
-export function calculateApiCostOpenAI(
-	modelInfo: ModelInfo,
-	inputTokens: number,
-	outputTokens: number,
-	cacheCreationInputTokens?: number,
-	cacheReadInputTokens?: number,
-): number {
-	const cacheCreationInputTokensNum = cacheCreationInputTokens || 0
-	const cacheReadInputTokensNum = cacheReadInputTokens || 0
-	const nonCachedInputTokens = Math.max(0, inputTokens - cacheCreationInputTokensNum - cacheReadInputTokensNum)
-	return calculateApiCostInternal(
-		modelInfo,
-		nonCachedInputTokens,
-		outputTokens,
-		cacheCreationInputTokensNum,
-		cacheReadInputTokensNum,
-	)
-}

src/utils/time.ts

@@ -0,0 +1,3 @@
+export function secondsToMs(seconds: number): number {
+	return seconds * 1000
+}

webview-ui/src/components/mcp/McpView.tsx

@@ -1,7 +1,15 @@
-import { VSCodeButton, VSCodeLink, VSCodePanels, VSCodePanelTab, VSCodePanelView } from "@vscode/webview-ui-toolkit/react"
+import {
+	VSCodeButton,
+	VSCodeLink,
+	VSCodePanels,
+	VSCodePanelTab,
+	VSCodePanelView,
+	VSCodeDropdown,
+	VSCodeOption,
+} from "@vscode/webview-ui-toolkit/react"
 import { useEffect, useState } from "react"
 import styled from "styled-components"
-import { McpServer } from "../../../../src/shared/mcp"
+import { DEFAULT_MCP_TIMEOUT_SECONDS, McpServer } from "../../../../src/shared/mcp"
 import { useExtensionState } from "../../context/ExtensionStateContext"
 import { getMcpServerDisplayName } from "../../utils/mcp"
 import { vscode } from "../../utils/vscode"
@@ -210,6 +218,36 @@ const ServerRow = ({ server }: { server: McpServer }) => {
 		}
 	}
 
+	const [timeoutValue, setTimeoutValue] = useState<string>(() => {
+		try {
+			const config = JSON.parse(server.config)
+			return config.timeout?.toString() || DEFAULT_MCP_TIMEOUT_SECONDS.toString()
+		} catch {
+			return DEFAULT_MCP_TIMEOUT_SECONDS.toString()
+		}
+	})
+
+	const timeoutOptions = [
+		{ value: "30", label: "30 seconds" },
+		{ value: "60", label: "1 minute" },
+		{ value: "300", label: "5 minutes" },
+		{ value: "600", label: "10 minutes" },
+		{ value: "1800", label: "30 minutes" },
+		{ value: "3600", label: "1 hour" },
+	]
+
+	const handleTimeoutChange = (e: any) => {
+		const select = e.target as HTMLSelectElement
+		const value = select.value
+		const num = parseInt(value)
+		setTimeoutValue(value)
+		vscode.postMessage({
+			type: "updateMcpTimeout",
+			serverName: server.name,
+			timeout: num,
+		})
+	}
+
 	const handleRestart = () => {
 		vscode.postMessage({
 			type: "restartMcpServer",
@@ -410,6 +448,16 @@ const ServerRow = ({ server }: { server: McpServer }) => {
 							</VSCodePanelView>
 						</VSCodePanels>
 
+						<div style={{ margin: "10px 7px" }}>
+							<label style={{ display: "block", marginBottom: "4px", fontSize: "13px" }}>Request Timeout</label>
+							<VSCodeDropdown style={{ width: "100%" }} value={timeoutValue} onChange={handleTimeoutChange}>
+								{timeoutOptions.map((option) => (
+									<VSCodeOption key={option.value} value={option.value}>
+										{option.label}
+									</VSCodeOption>
+								))}
+							</VSCodeDropdown>
+						</div>
 						<VSCodeButton
 							appearance="secondary"
 							onClick={handleRestart}

webview-ui/src/components/settings/SettingsView.tsx

@@ -18,7 +18,7 @@ const SettingsView = ({ onDone }: SettingsViewProps) => {
 		customInstructions,
 		setCustomInstructions,
 		openRouterModels,
-    requestyModels,
+		requestyModels,
 		telemetrySetting,
 		setTelemetrySetting,
 	} = useExtensionState()