[{"_1":2,"_294":-5,"_295":-5},"loaderData",{"_3":4,"_34":35,"_36":37,"_241":242},"root",{"_5":6,"_22":23},"toasts",[7,14,16,18,20],{"_8":9,"_10":11,"_12":13},"uid","881c6be9-46fa-4379-ab68-16d8082fe974","title","This MCP server does not support inspection","type","error",{"_8":15,"_10":11,"_12":13},"c7c0823a-4e03-467e-85bc-3267ca8eb756",{"_8":17,"_10":11,"_12":13},"f28b55c3-f51d-4f45-9ec9-db25ce6323c1",{"_8":19,"_10":11,"_12":13},"02128b8d-5ae5-4b78-ab18-b3b42a296543",{"_8":21,"_10":11,"_12":13},"2a9317cc-49aa-46e6-95db-a98cf397bd05","visitor",{"_24":25,"_26":27},"viewportWidth",1280,"visitorSession",{"_28":29,"_31":-5,"_8":32,"_33":-5},"attributes",[30],"BOT","membership","FkO6lxsamu","userAccount","routes/_public/_layout",{},"routes/_public/mcp/servers/~namespace/~slug/_pages/_layout",{"_38":39,"_80":-5,"_81":40,"_41":82,"_159":160,"_240":135},"availableLocales",[40,52,59,66,73],{"_41":42,"_45":49,"_50":51},"mcpServer",{"_43":44,"_45":46,"_47":48},"description","Generate an MCP server for any OpenAPI documented endpoint.","name","Emcee","readme","![emcee flow diagram](https://github.com/user-attachments/assets/bcac98a5-497f-4b34-9e8d-d4bc08852ea1)\n\n# emcee\n\n**emcee** is a tool that provides a [Model Context Protocol (MCP)][mcp] server\nfor any web application with an [OpenAPI][openapi] specification.\nYou can use emcee to connect [Claude Desktop][claude] and [other apps][mcp-clients]\nto external tools and data services,\nsimilar to [ChatGPT plugins][chatgpt-plugins].\n\n## Quickstart\n\nIf you're on macOS and have [Homebrew][homebrew] installed,\nyou can get up-and-running quickly.\n\n```bash\n# Install emcee\nbrew install loopwork-ai/tap/emcee\n```\n\nMake sure you have [Claude Desktop](https://claude.ai/download) installed.\n\nTo configure Claude Desktop for use with emcee:\n\n1. Open Claude Desktop Settings (⌘,)\n2. Select the \"Developer\" section in the sidebar\n3. Click \"Edit Config\" to open the configuration file\n\n![Claude Desktop settings Edit Config button](https://github.com/user-attachments/assets/761c6de5-62c2-4c53-83e6-54362040acd5)\n\nThe configuration file should be located in the Application Support directory.\nYou can also open it directly in VSCode using:\n\n```console\ncode ~/Library/Application\\ Support/Claude/claude_desktop_config.json\n```\n\nAdd the following configuration to add the weather.gov MCP server:\n\n```json\n{\n \"mcpServers\": {\n \"weather\": {\n \"command\": \"emcee\",\n \"args\": [\n \"https://api.weather.gov/openapi.json\"\n ]\n }\n }\n}\n```\n\nAfter saving the file, quit and re-open Claude.\nYou should now see 🔨57 in the bottom right corner of your chat box.\nClick on that to see a list of all the tools made available to Claude through MCP.\n\nStart a new chat and ask it about the weather where you are.\n\n> What's the weather in Portland, OR?\n\nClaude will consult the tools made available to it through MCP\nand request to use one if deemed to be suitable for answering your question.\nYou can review this request and either approve or deny it.\n\n $\"Allow$ \n\nIf you allow, Claude will communicate with the MCP\nand use the result to inform its response.\n\n![Claude response with MCP tool use](https://github.com/user-attachments/assets/d5b63002-1ed3-4b03-bc71-8357427bb06b)\n\n---\n\n> [!TIP]\n> Building agents? Want to deploy remote MCP servers?\n> Reach out to us at emcee@loopwork.com\n\n---\n\n## Why use emcee?\n\nMCP provides a standardized way to connect AI models to tools and data sources.\nIt's still early days, but there are already a variety of [available servers][mcp-servers]\nfor connecting to browsers, developer tools, and other systems.\n\nWe think emcee is a convenient way to connect to services\nthat don't have an existing MCP server implementation —\n_especially for services you're building yourself_.\nGot a web app with an OpenAPI spec?\nYou might be surprised how far you can get\nwithout a dashboard or client library.\n\n## Installation\n\n### Installer Script\n\nUse the [installer script][installer] to download and install a \n[pre-built release][releases] of emcee for your platform \n(Linux x86-64/i386/arm64 and macOS Intel/Apple Silicon).\n\n```console\n# fish\nsh (curl -fsSL https://get.emcee.sh | psub)\n\n# bash, zsh\nsh <(curl -fsSL https://get.emcee.sh)\n```\n\n### Homebrew\n\nInstall emcee using [Homebrew][homebrew] from [Loopwork's tap][homebrew-tap].\n\n```console\nbrew install loopwork-ai/tap/emcee\n```\n\n### Docker\n\nPrebuilt [Docker images][docker-images] with emcee are available.\n\n```console\ndocker run -it ghcr.io/loopwork-ai/emcee\n```\n\n### Build From Source\n\nRequires [go 1.24][golang] or later.\n\n```console\ngit clone https://github.com/loopwork-ai/emcee.git\ncd emcee\ngo build -o emcee cmd/emcee/main.go\n```\n\nOnce built, you can run in place (`./emcee`)\nor move it somewhere in your `PATH`, like `/usr/local/bin`.\n\n## Usage\n\n```console\nUsage:\n emcee [spec-path-or-url] [flags]\n\nFlags:\n --basic-auth string Basic auth value (either user:pass or base64 encoded, will be prefixed with 'Basic ')\n --bearer-auth string Bearer token value (will be prefixed with 'Bearer ')\n -h, --help help for emcee\n --raw-auth string Raw value for Authorization header\n --retries int Maximum number of retries for failed requests (default 3)\n -r, --rps int Maximum requests per second (0 for no limit)\n -s, --silent Disable all logging\n --timeout duration HTTP request timeout (default 1m0s)\n -v, --verbose Enable debug level logging to stderr\n --version version for emcee\n```\n\nemcee implements [Standard Input/Output (stdio)](https://modelcontextprotocol.io/docs/concepts/transports#standard-input-output-stdio) transport for MCP,\nwhich uses [JSON-RPC 2.0](https://www.jsonrpc.org/) as its wire format.\n\nWhen you run emcee from the command-line,\nit starts a program that listens on stdin,\noutputs to stdout,\nand logs to stderr.\n\n### Authentication\n\nFor APIs that require authentication,\nemcee supports several authentication methods:\n\n| Authentication Type | Example Usage | Resulting Header |\n|------------------------|---------------|----------------------------|\n| **Bearer Token** | `--bearer-auth=\"abc123\"` | `Authorization: Bearer abc123` |\n| **Basic Auth** | `--basic-auth=\"user:pass\"` | `Authorization: Basic dXNlcjpwYXNz` |\n| **Raw Value** | `--raw-auth=\"Custom xyz789\"` | `Authorization: Custom xyz789` |\n\nThese authentication values can be provided directly\nor as [1Password secret references][secret-reference-syntax].\n\nWhen using 1Password references:\n- Use the format `op://vault/item/field`\n (e.g. `--bearer-auth=\"op://Shared/X/credential\"`)\n- Ensure the 1Password CLI ([op][op]) is installed and available in your `PATH`\n- Sign in to 1Password before running emcee or launching Claude Desktop\n\n```console\n# Install op\nbrew install 1password-cli\n\n# Sign in 1Password CLI\nop signin\n```\n\n```json\n{\n \"mcpServers\": {\n \"twitter\": {\n \"command\": \"emcee\",\n \"args\": [\n \"--bearer-auth=op://shared/x/credential\",\n \"https://api.twitter.com/2/openapi.json\"\n ]\n }\n }\n}\n```\n\n $\"1Password$ \n\n> [!IMPORTANT] \n> emcee doesn't use auth credentials when downloading \n> OpenAPI specifications from URLs provided as command arguments.\n> If your OpenAPI specification requires authentication to access,\n> first download it to a local file using your preferred HTTP client,\n> then provide the local file path to emcee.\n\n### Transforming OpenAPI Specifications\n\nYou can transform OpenAPI specifications before passing them to emcee using standard Unix utilities. This is useful for:\n- Selecting specific endpoints to expose as tools\n with [jq][jq] or [yq][yq]\n- Modifying descriptions or parameters \n with [OpenAPI Overlays][openapi-overlays]\n- Combining multiple specifications\n with [Redocly][redocly-cli]\n\nFor example,\nyou can use `jq` to include only the `point` tool from `weather.gov`.\n\n```console\ncat path/to/openapi.json | \\\n jq 'if .paths then .paths |= with_entries(select(.key == \"/points/{point}\")) else . end' | \\\n emcee\n```\n\n\n### JSON-RPC\n\nYou can interact directly with the provided MCP server\nby sending JSON-RPC requests.\n\n> [!NOTE]\n> emcee provides only MCP tool capabilities.\n> Other features like resources, prompts, and sampling aren't yet supported.\n\n#### List Tools\n\n

\n\n

Request

\n\n```json\n{\"jsonrpc\": \"2.0\", \"method\": \"tools/list\", \"params\": {}, \"id\": 1}\n```\n\n

\n\n

Response

\n\n```jsonc\n{\n \"jsonrpc\":\"2.0\",\n \"result\": {\n \"tools\": [\n // ...\n {\n \"name\": \"tafs\",\n \"description\": \"Returns Terminal Aerodrome Forecasts for the specified airport station.\",\n \"inputSchema\": {\n \"type\": \"object\",\n \"properties\": {\n \"stationId\": {\n \"description\": \"Observation station ID\",\n \"type\": \"string\"\n }\n },\n \"required\": [\"stationId\"]\n }\n },\n // ...\n ]\n },\n \"id\": 1\n}\n```\n

\n\n#### Call Tool\n\n

\n\n

Request

\n\n```json\n{\"jsonrpc\": \"2.0\", \"method\": \"tools/call\", \"params\": { \"name\": \"taf\", \"arguments\": { \"stationId\": \"KPDX\" }}, \"id\": 1}\n```\n\n

\n\n

Response

\n\n```jsonc\n{\n \"jsonrpc\":\"2.0\",\n \"content\": [\n {\n \"type\": \"text\",\n \"text\": /* Weather forecast in GeoJSON format */,\n \"annotations\": {\n \"audience\": [\"assistant\"]\n }\n }\n ]\n \"id\": 1\n}\n```\n\n

\n\n## Debugging\n\nThe [MCP Inspector][mcp-inspector] is a tool for testing and debugging MCP servers.\nIf Claude and/or emcee aren't working as expected,\nthe inspector can help you understand what's happening.\n\n```console\nnpx @modelcontextprotocol/inspector emcee https://api.weather.gov/openapi.json\n# 🔍 MCP Inspector is up and running at http://localhost:5173 🚀\n```\n\n```console\nopen http://localhost:5173\n```\n\n## License\n\nemcee is licensed under the Apache License, Version 2.0.\n\n[chatgpt-plugins]: https://openai.com/index/chatgpt-plugins/\n[claude]: https://claude.ai/download\n[docker-images]: https://github.com/loopwork-ai/emcee/pkgs/container/emcee\n[golang]: https://go.dev\n[homebrew]: https://brew.sh\n[homebrew-tap]: https://github.com/loopwork-ai/homebrew-tap\n[installer]: https://github.com/loopwork-ai/emcee/blob/main/tools/install.sh\n[jq]: https://github.com/jqlang/jq\n[mcp]: https://modelcontextprotocol.io/\n[mcp-clients]: https://modelcontextprotocol.info/docs/clients/\n[mcp-inspector]: https://github.com/modelcontextprotocol/inspector\n[mcp-servers]: https://modelcontextprotocol.io/examples\n[op]: https://developer.1password.com/docs/cli/get-started/\n[openapi]: https://openapi.org\n[openapi-overlays]: https://www.openapis.org/blog/2024/10/22/announcing-overlay-specification\n[redocly-cli]: https://redocly.com/docs/cli/commands\n[releases]: https://github.com/loopwork-ai/emcee/releases\n[secret-reference-syntax]: https://developer.1password.com/docs/cli/secret-reference-syntax/\n[yq]: https://github.com/mikefarah/yq","English","tag","en-US",{"_41":53,"_45":57,"_50":58},{"_43":54,"_45":55,"_47":56},"Genere un servidor MCP para cualquier punto final documentado de OpenAPI.","Maestro de ceremonias","![diagrama de flujo del maestro de ceremonias](https://github.com/user-attachments/assets/bcac98a5-497f-4b34-9e8d-d4bc08852ea1)\n\n# maestro de ceremonias\n\n**emcee** es una herramienta que proporciona un servidor [de Protocolo de Contexto de Modelo (MCP)](https://modelcontextprotocol.io/) para cualquier aplicación web con especificación [OpenAPI](https://openapi.org) . Puedes usar emcee para conectar [Claude Desktop](https://claude.ai/download) y [otras aplicaciones](https://modelcontextprotocol.info/docs/clients/) a herramientas y servicios de datos externos, de forma similar a [los plugins de ChatGPT](https://openai.com/index/chatgpt-plugins/) .\n\n## Inicio rápido\n\nSi estás en macOS y tienes [Homebrew](https://brew.sh) instalado, podrás empezar a usarlo rápidamente.\n\n```bash\n# Install emcee\nbrew install loopwork-ai/tap/emcee\n```\n\nAsegúrese de tener instalado [Claude Desktop](https://claude.ai/download) .\n\nPara configurar Claude Desktop para su uso con emcee:\n\n1. Abra la configuración del escritorio de Claude (⌘,)\n2. Seleccione la sección \"Desarrollador\" en la barra lateral.\n3. Haga clic en \"Editar configuración\" para abrir el archivo de configuración\n\n![Botón Editar configuración de Claude Desktop](https://github.com/user-attachments/assets/761c6de5-62c2-4c53-83e6-54362040acd5)\n\nEl archivo de configuración debe estar ubicado en el directorio de soporte de la aplicación. También puede abrirlo directamente en VSCode usando:\n\n```console\ncode ~/Library/Application\\ Support/Claude/claude_desktop_config.json\n```\n\nAgregue la siguiente configuración para agregar el servidor MCP de weather.gov:\n\n```json\n{\n \"mcpServers\": {\n \"weather\": {\n \"command\": \"emcee\",\n \"args\": [\n \"https://api.weather.gov/openapi.json\"\n ]\n }\n }\n}\n```\n\nDespués de guardar el archivo, cierra y vuelve a abrir Claude. Deberías ver 🔨57 en la esquina inferior derecha del chat. Haz clic en él para ver una lista de todas las herramientas disponibles para Claude a través de MCP.\n\nInicia un nuevo chat y pregúntale sobre el clima donde estás.\n\n> ¿Cómo es el clima en Portland, OR?\n\nClaude consultará las herramientas disponibles a través de MCP y solicitará usar una si la considera adecuada para responder a su pregunta. Puede revisar esta solicitud y aprobarla o rechazarla.\n\nSi lo permite, Claude se comunicará con el MCP y utilizará el resultado para informar su respuesta.\n\n![Respuesta de Claude con el uso de la herramienta MCP](https://github.com/user-attachments/assets/d5b63002-1ed3-4b03-bc71-8357427bb06b)\n\n***\n\n> \\[!TIP] ¿Creando agentes? ¿Quieres implementar servidores MCP remotos? Contáctanos en \n\n***\n\n## ¿Por qué utilizar emcee?\n\nMCP proporciona una forma estandarizada de conectar modelos de IA a herramientas y fuentes de datos. Aún es pronto, pero ya existen diversos [servidores disponibles](https://modelcontextprotocol.io/examples) para conectarse a navegadores, herramientas de desarrollo y otros sistemas.\n\nCreemos que emcee es una forma práctica de conectar servicios que no cuentan con una implementación de servidor MCP, *especialmente para servicios que estás desarrollando tú mismo* . ¿Tienes una aplicación web con una especificación OpenAPI? Te sorprenderá lo mucho que puedes llegar sin un panel de control ni una biblioteca cliente.\n\n## Instalación\n\n### Script de instalación\n\nUtilice el [script de instalación](https://github.com/loopwork-ai/emcee/blob/main/tools/install.sh) para descargar e instalar una [versión precompilada](https://github.com/loopwork-ai/emcee/releases) de emcee para su plataforma (Linux x86-64/i386/arm64 y macOS Intel/Apple Silicon).\n\n```console\n# fish\nsh (curl -fsSL https://get.emcee.sh | psub)\n\n# bash, zsh\nsh <(curl -fsSL https://get.emcee.sh)\n```\n\n### Cerveza casera\n\nInstalar emcee usando [Homebrew](https://brew.sh) desde [el grifo de Loopwork](https://github.com/loopwork-ai/homebrew-tap) .\n\n```console\nbrew install loopwork-ai/tap/emcee\n```\n\n### Estibador\n\nHay disponibles [imágenes de Docker](https://github.com/loopwork-ai/emcee/pkgs/container/emcee) prediseñadas con emcee.\n\n```console\ndocker run -it ghcr.io/loopwork-ai/emcee\n```\n\n### Construir desde la fuente\n\nRequiere [go 1.24](https://go.dev) o posterior.\n\n```console\ngit clone https://github.com/loopwork-ai/emcee.git\ncd emcee\ngo build -o emcee cmd/emcee/main.go\n```\n\nUna vez creado, puedes ejecutarlo en el lugar ( `./emcee` ) o moverlo a algún lugar en tu `PATH` , como `/usr/local/bin` .\n\n## Uso\n\n```console\nUsage:\n emcee [spec-path-or-url] [flags]\n\nFlags:\n --basic-auth string Basic auth value (either user:pass or base64 encoded, will be prefixed with 'Basic ')\n --bearer-auth string Bearer token value (will be prefixed with 'Bearer ')\n -h, --help help for emcee\n --raw-auth string Raw value for Authorization header\n --retries int Maximum number of retries for failed requests (default 3)\n -r, --rps int Maximum requests per second (0 for no limit)\n -s, --silent Disable all logging\n --timeout duration HTTP request timeout (default 1m0s)\n -v, --verbose Enable debug level logging to stderr\n --version version for emcee\n```\n\nemcee implementa el transporte [de entrada/salida estándar (stdio)](https://modelcontextprotocol.io/docs/concepts/transports#standard-input-output-stdio) para MCP, que utiliza [JSON-RPC 2.0](https://www.jsonrpc.org/) como su formato de cable.\n\nCuando ejecuta emcee desde la línea de comandos, inicia un programa que escucha en stdin, genera salida en stdout y registra en stderr.\n\n### Autenticación\n\nPara las API que requieren autenticación, emcee admite varios métodos de autenticación:\n\n| Tipo de autenticación | Ejemplo de uso | Encabezado resultante |\n| ----------------------- | ---------------------------- | ----------------------------------- |\n| **Token al portador** | `--bearer-auth=\"abc123\"` | `Authorization: Bearer abc123` |\n| **Autorización básica** | `--basic-auth=\"user:pass\"` | `Authorization: Basic dXNlcjpwYXNz` |\n| **Valor bruto** | `--raw-auth=\"Custom xyz789\"` | `Authorization: Custom xyz789` |\n\nEstos valores de autenticación se pueden proporcionar directamente o como [referencias secretas de 1Password](https://developer.1password.com/docs/cli/secret-reference-syntax/) .\n\nAl utilizar referencias de 1Password:\n\n* Utilice el formato `op://vault/item/field` (por ejemplo, `--bearer-auth=\"op://Shared/X/credential\"` )\n* Asegúrese de que la CLI de 1Password ( [op](https://developer.1password.com/docs/cli/get-started/) ) esté instalada y disponible en su `PATH`\n* Inicie sesión en 1Password antes de ejecutar emcee o iniciar Claude Desktop\n\n```console\n# Install op\nbrew install 1password-cli\n\n# Sign in 1Password CLI\nop signin\n```\n\n```json\n{\n \"mcpServers\": {\n \"twitter\": {\n \"command\": \"emcee\",\n \"args\": [\n \"--bearer-auth=op://shared/x/credential\",\n \"https://api.twitter.com/2/openapi.json\"\n ]\n }\n }\n}\n```\n\n> \\[!IMPORTANTE]\\\n> emcee no usa credenciales de autenticación al descargar especificaciones de OpenAPI desde las URL proporcionadas como argumentos de comando. Si su especificación de OpenAPI requiere autenticación para acceder, primero descárguela a un archivo local con su cliente HTTP preferido y luego proporcione la ruta del archivo local a emcee.\n\n### Transformación de las especificaciones de OpenAPI\n\nPuedes transformar las especificaciones de OpenAPI antes de pasarlas a emcee mediante utilidades estándar de Unix. Esto es útil para:\n\n* Seleccionar puntos finales específicos para exponer como herramientas con [jq](https://github.com/jqlang/jq) o [yq](https://github.com/mikefarah/yq)\n* Modificar descripciones o parámetros con [superposiciones de OpenAPI](https://www.openapis.org/blog/2024/10/22/announcing-overlay-specification)\n* Combinando múltiples especificaciones con [Redocly](https://redocly.com/docs/cli/commands)\n\nPor ejemplo, puede utilizar `jq` para incluir solo la herramienta `point` de `weather.gov` .\n\n```console\ncat path/to/openapi.json | \\\n jq 'if .paths then .paths |= with_entries(select(.key == \"/points/{point}\")) else . end' | \\\n emcee\n```\n\n### JSON-RPC\n\nPuede interactuar directamente con el servidor MCP proporcionado enviando solicitudes JSON-RPC.\n\n> \\[!NOTA] emcee solo ofrece funciones de la herramienta MCP. Otras funciones, como recursos, indicaciones y ejemplos, aún no son compatibles.\n\n#### Herramientas de lista\n\n```json\n{\"jsonrpc\": \"2.0\", \"method\": \"tools/list\", \"params\": {}, \"id\": 1}\n```\n\n```jsonc\n{\n \"jsonrpc\":\"2.0\",\n \"result\": {\n \"tools\": [\n // ...\n {\n \"name\": \"tafs\",\n \"description\": \"Returns Terminal Aerodrome Forecasts for the specified airport station.\",\n \"inputSchema\": {\n \"type\": \"object\",\n \"properties\": {\n \"stationId\": {\n \"description\": \"Observation station ID\",\n \"type\": \"string\"\n }\n },\n \"required\": [\"stationId\"]\n }\n },\n // ...\n ]\n },\n \"id\": 1\n}\n```\n\n#### Herramienta de llamada\n\n```json\n{\"jsonrpc\": \"2.0\", \"method\": \"tools/call\", \"params\": { \"name\": \"taf\", \"arguments\": { \"stationId\": \"KPDX\" }}, \"id\": 1}\n```\n\n```jsonc\n{\n \"jsonrpc\":\"2.0\",\n \"content\": [\n {\n \"type\": \"text\",\n \"text\": /* Weather forecast in GeoJSON format */,\n \"annotations\": {\n \"audience\": [\"assistant\"]\n }\n }\n ]\n \"id\": 1\n}\n```\n\n## Depuración\n\nEl [Inspector MCP](https://github.com/modelcontextprotocol/inspector) es una herramienta para probar y depurar servidores MCP. Si Claude o emcee no funcionan correctamente, el inspector puede ayudarle a comprender qué sucede.\n\n```console\nnpx @modelcontextprotocol/inspector emcee https://api.weather.gov/openapi.json\n# 🔍 MCP Inspector is up and running at http://localhost:5173 🚀\n```\n\n```console\nopen http://localhost:5173\n```\n\n## Licencia\n\nemcee tiene licencia Apache, versión 2.0.","Spanish","es-ES",{"_41":60,"_45":64,"_50":65},{"_43":61,"_45":62,"_47":63},"任意の OpenAPI ドキュメント化されたエンドポイント用の MCP サーバーを生成します。","司会者","![司会フロー図](https://github.com/user-attachments/assets/bcac98a5-497f-4b34-9e8d-d4bc08852ea1)\n\n# 司会者\n\n**emceeは**、 [OpenAPI](https://openapi.org)仕様に準拠したあらゆるWebアプリケーションに[Model Context Protocol（MCP）](https://modelcontextprotocol.io/)サーバーを提供するツールです。emceeを使用すると、 [Claude Desktop](https://claude.ai/download)や[その他のアプリを](https://modelcontextprotocol.info/docs/clients/)[ChatGPTプラグイン](https://openai.com/index/chatgpt-plugins/)と同様に外部ツールやデータサービスに接続できます。\n\n## クイックスタート\n\nmacOS を使用していて[Homebrew](https://brew.sh)がインストールされている場合は、すぐに起動して実行できます。\n\n```bash\n# Install emcee\nbrew install loopwork-ai/tap/emcee\n```\n\n[Claude Desktop](https://claude.ai/download)がインストールされていることを確認してください。\n\nClaude Desktop を emcee で使用するように構成するには:\n\n1. Claude デスクトップ設定を開く (⌘,)\n2. サイドバーの「開発者」セクションを選択します\n3. 「設定の編集」をクリックして設定ファイルを開きます\n\n![クロードデスクトップ設定編集設定ボタン](https://github.com/user-attachments/assets/761c6de5-62c2-4c53-83e6-54362040acd5)\n\n設定ファイルはApplication Supportディレクトリに保存されています。以下のコマンドでVSCodeから直接開くこともできます。\n\n```console\ncode ~/Library/Application\\ Support/Claude/claude_desktop_config.json\n```\n\nweather.gov MCP サーバーを追加するには、次の構成を追加します。\n\n```json\n{\n \"mcpServers\": {\n \"weather\": {\n \"command\": \"emcee\",\n \"args\": [\n \"https://api.weather.gov/openapi.json\"\n ]\n }\n }\n}\n```\n\nファイルを保存したら、Claudeを終了して再起動してください。チャットボックスの右下に🔨57が表示されます。これをクリックすると、MCPを通じてClaudeが利用できるすべてのツールのリストが表示されます。\n\n新しいチャットを開始して、現在地の天気について質問します。\n\n> オレゴン州ポートランドの天気はどうですか?\n\nClaudeはMCPを通じて利用可能なツールを確認し、お客様の質問への回答に適していると判断した場合は、そのツールの使用をリクエストします。このリクエストを確認し、承認または拒否することができます。\n\n許可すると、Claude は MCP と通信し、その結果を使用して応答を通知します。\n\n![MCPツールの使用に関するクロードの応答](https://github.com/user-attachments/assets/d5b63002-1ed3-4b03-bc71-8357427bb06b)\n\n***\n\n> \\[!TIP] エージェントを構築していますか？リモートMCPサーバーを展開したいですか[？emcee@loopwork.com](mailto:emcee@loopwork.com)までお問い合わせください。\n\n***\n\n## なぜ司会者を使うのですか?\n\nMCPは、AIモデルをツールやデータソースに接続するための標準化された方法を提供します。まだ初期段階ですが、ブラウザ、開発者ツール、その他のシステムに接続するためのさまざまな[サーバーが既に利用可能](https://modelcontextprotocol.io/examples)です。\n\nemceeは、既存のMCPサーバー実装がないサービス、*特に自分で構築しているサービスに接続するため*の便利な手段だと考えています。OpenAPI仕様のWebアプリをお持ちですか？ダッシュボードやクライアントライブラリなしで、ここまでできることに驚かれるかもしれません。\n\n## インストール\n\n### インストーラースクリプト\n\n[インストーラースクリプト](https://github.com/loopwork-ai/emcee/blob/main/tools/install.sh)を使用して、プラットフォーム (Linux x86-64/i386/arm64 および macOS Intel/Apple Silicon) 用の emcee の[ビルド済みリリース](https://github.com/loopwork-ai/emcee/releases)をダウンロードしてインストールします。\n\n```console\n# fish\nsh (curl -fsSL https://get.emcee.sh | psub)\n\n# bash, zsh\nsh <(curl -fsSL https://get.emcee.sh)\n```\n\n### 自家製ビール\n\n[Loopwork の tap](https://github.com/loopwork-ai/homebrew-tap)から[Homebrew](https://brew.sh)を使用して emcee をインストールします。\n\n```console\nbrew install loopwork-ai/tap/emcee\n```\n\n### ドッカー\n\nemcee を使用した事前構築済みの[Docker イメージが](https://github.com/loopwork-ai/emcee/pkgs/container/emcee)利用可能です。\n\n```console\ndocker run -it ghcr.io/loopwork-ai/emcee\n```\n\n### ソースからビルド\n\n[go 1.24](https://go.dev)以降が必要です。\n\n```console\ngit clone https://github.com/loopwork-ai/emcee.git\ncd emcee\ngo build -o emcee cmd/emcee/main.go\n```\n\nビルドしたら、その場で実行 ( `./emcee` ) することも、 `/usr/local/bin`など`PATH`内のどこかに移動することもできます。\n\n## 使用法\n\n```console\nUsage:\n emcee [spec-path-or-url] [flags]\n\nFlags:\n --basic-auth string Basic auth value (either user:pass or base64 encoded, will be prefixed with 'Basic ')\n --bearer-auth string Bearer token value (will be prefixed with 'Bearer ')\n -h, --help help for emcee\n --raw-auth string Raw value for Authorization header\n --retries int Maximum number of retries for failed requests (default 3)\n -r, --rps int Maximum requests per second (0 for no limit)\n -s, --silent Disable all logging\n --timeout duration HTTP request timeout (default 1m0s)\n -v, --verbose Enable debug level logging to stderr\n --version version for emcee\n```\n\nemcee は、ワイヤ形式として[JSON-RPC 2.0 を](https://www.jsonrpc.org/)使用する MCP 用の[標準入出力 (stdio)](https://modelcontextprotocol.io/docs/concepts/transports#standard-input-output-stdio)トランスポートを実装します。\n\nコマンドラインから emcee を実行すると、stdin をリッスンし、stdout に出力し、stderr にログを記録するプログラムが起動します。\n\n### 認証\n\n認証を必要とする API の場合、emcee はいくつかの認証方法をサポートしています。\n\n| 認証タイプ | 使用例 | 結果のヘッダー |\n| ----------- | ---------------------------- | ----------------------------------- |\n| **無記名トークン** | `--bearer-auth=\"abc123\"` | `Authorization: Bearer abc123` |\n| **基本認証** | `--basic-auth=\"user:pass\"` | `Authorization: Basic dXNlcjpwYXNz` |\n| **生の値** | `--raw-auth=\"Custom xyz789\"` | `Authorization: Custom xyz789` |\n\nこれらの認証値は直接提供することも、 [1Password シークレット参照](https://developer.1password.com/docs/cli/secret-reference-syntax/)として提供することもできます。\n\n1Password 参照を使用する場合:\n\n* `op://vault/item/field`形式を使用します (例`--bearer-auth=\"op://Shared/X/credential\"` )\n* 1Password CLI ( [op](https://developer.1password.com/docs/cli/get-started/) )がインストールされ、 `PATH`で利用可能であることを確認してください。\n* emcee を実行する前、または Claude Desktop を起動する前に 1Password にサインインしてください\n\n```console\n# Install op\nbrew install 1password-cli\n\n# Sign in 1Password CLI\nop signin\n```\n\n```json\n{\n \"mcpServers\": {\n \"twitter\": {\n \"command\": \"emcee\",\n \"args\": [\n \"--bearer-auth=op://shared/x/credential\",\n \"https://api.twitter.com/2/openapi.json\"\n ]\n }\n }\n}\n```\n\n> \\[！重要]\\\n> emcee は、コマンド引数として指定された URL から OpenAPI 仕様をダウンロードする際に認証情報を使用しません。OpenAPI 仕様へのアクセスに認証が必要な場合は、まずお好みの HTTP クライアントを使用してローカルファイルにダウンロードし、そのローカルファイルのパスを emcee に指定してください。\n\n### OpenAPI仕様の変革\n\nOpenAPI仕様をemceeに渡す前に、標準のUnixユーティリティを使用して変換することができます。これは以下の場合に便利です。\n\n* [jq](https://github.com/jqlang/jq)または[yq](https://github.com/mikefarah/yq)を使用してツールとして公開する特定のエンドポイントを選択する\n* [OpenAPI オーバーレイ](https://www.openapis.org/blog/2024/10/22/announcing-overlay-specification)を使用して説明またはパラメータを変更する\n* [Redocly](https://redocly.com/docs/cli/commands)で複数の仕様を組み合わせる\n\nたとえば、 `jq`を使用して`weather.gov`の`point`ツールのみを含めることができます。\n\n```console\ncat path/to/openapi.json | \\\n jq 'if .paths then .paths |= with_entries(select(.key == \"/points/{point}\")) else . end' | \\\n emcee\n```\n\n### JSON-RPC\n\nJSON-RPC リクエストを送信することで、提供されている MCP サーバーと直接対話できます。\n\n> \\[!NOTE] emcee は MCP ツールの機能のみを提供します。リソース、プロンプト、サンプリングなどのその他の機能はまだサポートされていません。\n\n#### リストツール\n\n```json\n{\"jsonrpc\": \"2.0\", \"method\": \"tools/list\", \"params\": {}, \"id\": 1}\n```\n\n```jsonc\n{\n \"jsonrpc\":\"2.0\",\n \"result\": {\n \"tools\": [\n // ...\n {\n \"name\": \"tafs\",\n \"description\": \"Returns Terminal Aerodrome Forecasts for the specified airport station.\",\n \"inputSchema\": {\n \"type\": \"object\",\n \"properties\": {\n \"stationId\": {\n \"description\": \"Observation station ID\",\n \"type\": \"string\"\n }\n },\n \"required\": [\"stationId\"]\n }\n },\n // ...\n ]\n },\n \"id\": 1\n}\n```\n\n#### 通話ツール\n\n```json\n{\"jsonrpc\": \"2.0\", \"method\": \"tools/call\", \"params\": { \"name\": \"taf\", \"arguments\": { \"stationId\": \"KPDX\" }}, \"id\": 1}\n```\n\n```jsonc\n{\n \"jsonrpc\":\"2.0\",\n \"content\": [\n {\n \"type\": \"text\",\n \"text\": /* Weather forecast in GeoJSON format */,\n \"annotations\": {\n \"audience\": [\"assistant\"]\n }\n }\n ]\n \"id\": 1\n}\n```\n\n## デバッグ\n\n[MCP Inspectorは](https://github.com/modelcontextprotocol/inspector)、MCPサーバーのテストとデバッグを行うツールです。Claudeやemceeが期待通りに動作しない場合は、このInspectorを使って原因を突き止めることができます。\n\n```console\nnpx @modelcontextprotocol/inspector emcee https://api.weather.gov/openapi.json\n# 🔍 MCP Inspector is up and running at http://localhost:5173 🚀\n```\n\n```console\nopen http://localhost:5173\n```\n\n## ライセンス\n\nemcee は、Apache License バージョン 2.0 に基づいてライセンスされます。","Japanese","ja-JP",{"_41":67,"_45":71,"_50":72},{"_43":68,"_45":69,"_47":70},"OpenAPI로 문서화된 모든 엔드포인트에 대한 MCP 서버를 생성합니다.","사회자","![MC 흐름도](https://github.com/user-attachments/assets/bcac98a5-497f-4b34-9e8d-d4bc08852ea1)\n\n# 사회자\n\n**emcee는** [OpenAPI](https://openapi.org) 사양을 사용하는 모든 웹 애플리케이션에 [모델 컨텍스트 프로토콜(MCP)](https://modelcontextprotocol.io/) 서버를 제공하는 도구입니다. emcee를 사용하면 [ChatGPT 플러그인](https://openai.com/index/chatgpt-plugins/) 처럼 [Claude Desktop](https://claude.ai/download) 및 [기타 앱을](https://modelcontextprotocol.info/docs/clients/) 외부 도구 및 데이터 서비스에 연결할 수 있습니다.\n\n## 빠른 시작\n\nmacOS를 사용하고 [Homebrew가](https://brew.sh) 설치되어 있다면 빠르게 작업을 시작할 수 있습니다.\n\n지엑스피1\n\n[Claude Desktop이](https://claude.ai/download) 설치되어 있는지 확인하세요.\n\nemcee와 함께 사용하도록 Claude Desktop을 구성하려면:\n\n1. Claude 데스크톱 설정 열기(⌘,)\n2. 사이드바에서 \"개발자\" 섹션을 선택하세요\n3. \"구성 편집\"을 클릭하여 구성 파일을 엽니다.\n\n![Claude Desktop 설정 구성 편집 버튼](https://github.com/user-attachments/assets/761c6de5-62c2-4c53-83e6-54362040acd5)\n\n구성 파일은 Application Support 디렉터리에 있어야 합니다. VSCode에서 다음 명령어를 사용하여 직접 열 수도 있습니다.\n\n```console\ncode ~/Library/Application\\ Support/Claude/claude_desktop_config.json\n```\n\nweather.gov MCP 서버를 추가하려면 다음 구성을 추가하세요.\n\n```json\n{\n \"mcpServers\": {\n \"weather\": {\n \"command\": \"emcee\",\n \"args\": [\n \"https://api.weather.gov/openapi.json\"\n ]\n }\n }\n}\n```\n\n파일을 저장한 후 Claude를 종료했다가 다시 열어 보세요. 채팅 상자 오른쪽 하단에 🔨57이 표시됩니다. 이 아이콘을 클릭하면 MCP를 통해 Claude가 사용할 수 있는 모든 도구 목록을 볼 수 있습니다.\n\n새로운 채팅을 시작하고 현재 위치의 날씨에 대해 물어보세요.\n\n> 오리건주 포틀랜드의 날씨는 어때요?\n\nClaude는 MCP를 통해 제공되는 도구를 검토하고, 귀하의 질문에 답변하는 데 적합하다고 판단되는 경우 해당 도구를 사용하도록 요청할 것입니다. 귀하는 이 요청을 검토하고 승인 또는 거부할 수 있습니다.\n\n허락하신다면, 클로드는 MCP와 소통하여 그 결과를 바탕으로 대응 방안을 알려드립니다.\n\n![MCP 도구 사용에 대한 Claude의 반응](https://github.com/user-attachments/assets/d5b63002-1ed3-4b03-bc71-8357427bb06b)\n\n***\n\n> \\[!TIP] 에이전트를 구축하고 계신가요? 원격 MCP 서버를 구축하고 싶으신가요? 으로 문의해 주세요.\n\n***\n\n## 왜 emcee를 사용하나요?\n\nMCP는 AI 모델을 도구 및 데이터 소스에 연결하는 표준화된 방법을 제공합니다. 아직 초기 단계이지만, 브라우저, 개발자 도구 및 기타 시스템에 연결할 수 [있는 다양한 서버가](https://modelcontextprotocol.io/examples) 이미 존재합니다.\n\nemcee는 기존 MCP 서버 구현이 없는 서비스, *특히 직접 구축하는 서비스* 에 연결하는 편리한 방법이라고 생각합니다. OpenAPI 사양을 준수하는 웹앱이 있으신가요? 대시보드나 클라이언트 라이브러리 없이도 얼마나 많은 것을 할 수 있는지 놀라실 겁니다.\n\n## 설치\n\n### 설치 프로그램 스크립트\n\n[설치 스크립트를](https://github.com/loopwork-ai/emcee/blob/main/tools/install.sh) 사용하여 플랫폼(Linux x86-64/i386/arm64 및 macOS Intel/Apple Silicon)에 맞게 [미리 빌드된 emcee 릴리스를](https://github.com/loopwork-ai/emcee/releases) 다운로드하고 설치합니다.\n\n```console\n# fish\nsh (curl -fsSL https://get.emcee.sh | psub)\n\n# bash, zsh\nsh <(curl -fsSL https://get.emcee.sh)\n```\n\n### 홈브루\n\n[Loopwork의 탭](https://github.com/loopwork-ai/homebrew-tap) 에서 [Homebrew를](https://brew.sh) 사용하여 emcee를 설치합니다.\n\n```console\nbrew install loopwork-ai/tap/emcee\n```\n\n### 도커\n\nemcee를 사용하여 미리 빌드된 [Docker 이미지를](https://github.com/loopwork-ai/emcee/pkgs/container/emcee) 사용할 수 있습니다.\n\n```console\ndocker run -it ghcr.io/loopwork-ai/emcee\n```\n\n### 소스에서 빌드\n\n[go 1.24](https://go.dev) 이상이 필요합니다.\n\n```console\ngit clone https://github.com/loopwork-ai/emcee.git\ncd emcee\ngo build -o emcee cmd/emcee/main.go\n```\n\n빌드한 후에는 해당 위치에서 실행할 수 있습니다( `./emcee` ) 또는 `PATH` 의 어딘가( `/usr/local/bin` 로 이동할 수 있습니다.\n\n## 용법\n\n```console\nUsage:\n emcee [spec-path-or-url] [flags]\n\nFlags:\n --basic-auth string Basic auth value (either user:pass or base64 encoded, will be prefixed with 'Basic ')\n --bearer-auth string Bearer token value (will be prefixed with 'Bearer ')\n -h, --help help for emcee\n --raw-auth string Raw value for Authorization header\n --retries int Maximum number of retries for failed requests (default 3)\n -r, --rps int Maximum requests per second (0 for no limit)\n -s, --silent Disable all logging\n --timeout duration HTTP request timeout (default 1m0s)\n -v, --verbose Enable debug level logging to stderr\n --version version for emcee\n```\n\nemcee는 MCP에 대한 [표준 입출력(stdio)](https://modelcontextprotocol.io/docs/concepts/transports#standard-input-output-stdio) 전송을 구현하는데, 이는 와이어 포맷으로 [JSON-RPC 2.0을](https://www.jsonrpc.org/) 사용합니다.\n\n명령줄에서 emcee를 실행하면 stdin에서 수신하고 stdout에 출력하고 stderr에 로깅하는 프로그램이 시작됩니다.\n\n### 입증\n\n인증이 필요한 API의 경우 emcee는 여러 인증 방법을 지원합니다.\n\n| 인증 유형 | 사용 예 | 결과 헤더 |\n| ---------- | ---------------------------- | ----------------------------------- |\n| **베어러 토큰** | `--bearer-auth=\"abc123\"` | `Authorization: Bearer abc123` |\n| **기본 인증** | `--basic-auth=\"user:pass\"` | `Authorization: Basic dXNlcjpwYXNz` |\n| **원시 가치** | `--raw-auth=\"Custom xyz789\"` | `Authorization: Custom xyz789` |\n\n이러한 인증 값은 직접 제공되거나 [1Password 비밀번호 참조](https://developer.1password.com/docs/cli/secret-reference-syntax/) 로 제공될 수 있습니다.\n\n1Password 참조를 사용할 때:\n\n* `op://vault/item/field` 형식을 사용하세요(예 `--bearer-auth=\"op://Shared/X/credential\"` )\n* 1Password CLI( [op](https://developer.1password.com/docs/cli/get-started/) )가 설치되어 `PATH` 에서 사용 가능한지 확인하세요.\n* emcee를 실행하거나 Claude Desktop을 시작하기 전에 1Password에 로그인하세요.\n\n```console\n# Install op\nbrew install 1password-cli\n\n# Sign in 1Password CLI\nop signin\n```\n\n```json\n{\n \"mcpServers\": {\n \"twitter\": {\n \"command\": \"emcee\",\n \"args\": [\n \"--bearer-auth=op://shared/x/credential\",\n \"https://api.twitter.com/2/openapi.json\"\n ]\n }\n }\n}\n```\n\n> \\[!중요한]\\\n> emcee는 명령 인수로 제공된 URL에서 OpenAPI 사양을 다운로드할 때 인증 자격 증명을 사용하지 않습니다. OpenAPI 사양에 액세스하기 위해 인증이 필요한 경우, 먼저 선호하는 HTTP 클라이언트를 사용하여 로컬 파일에 다운로드한 후 emcee에 로컬 파일 경로를 제공하세요.\n\n### OpenAPI 사양 변환\n\n표준 Unix 유틸리티를 사용하여 OpenAPI 사양을 emcee에 전달하기 전에 변환할 수 있습니다. 이 기능은 다음과 같은 경우에 유용합니다.\n\n* [jq](https://github.com/jqlang/jq) 또는 [yq를](https://github.com/mikefarah/yq) 사용하여 도구로 노출할 특정 엔드포인트 선택\n* [OpenAPI 오버레이를](https://www.openapis.org/blog/2024/10/22/announcing-overlay-specification) 사용하여 설명 또는 매개변수 수정\n* [Redocly를](https://redocly.com/docs/cli/commands) 사용하여 여러 사양 결합\n\n예를 들어, `jq` 사용하면 `weather.gov` 의 `point` 도구만 포함할 수 있습니다.\n\n```console\ncat path/to/openapi.json | \\\n jq 'if .paths then .paths |= with_entries(select(.key == \"/points/{point}\")) else . end' | \\\n emcee\n```\n\n### JSON-RPC\n\nJSON-RPC 요청을 보내면 제공된 MCP 서버와 직접 상호 작용할 수 있습니다.\n\n> \\[!NOTE] emcee는 MCP 도구 기능만 제공합니다. 리소스, 프롬프트, 샘플링과 같은 다른 기능은 아직 지원되지 않습니다.\n\n#### 목록 도구\n\n```json\n{\"jsonrpc\": \"2.0\", \"method\": \"tools/list\", \"params\": {}, \"id\": 1}\n```\n\n```jsonc\n{\n \"jsonrpc\":\"2.0\",\n \"result\": {\n \"tools\": [\n // ...\n {\n \"name\": \"tafs\",\n \"description\": \"Returns Terminal Aerodrome Forecasts for the specified airport station.\",\n \"inputSchema\": {\n \"type\": \"object\",\n \"properties\": {\n \"stationId\": {\n \"description\": \"Observation station ID\",\n \"type\": \"string\"\n }\n },\n \"required\": [\"stationId\"]\n }\n },\n // ...\n ]\n },\n \"id\": 1\n}\n```\n\n#### 통화 도구\n\n```json\n{\"jsonrpc\": \"2.0\", \"method\": \"tools/call\", \"params\": { \"name\": \"taf\", \"arguments\": { \"stationId\": \"KPDX\" }}, \"id\": 1}\n```\n\n```jsonc\n{\n \"jsonrpc\":\"2.0\",\n \"content\": [\n {\n \"type\": \"text\",\n \"text\": /* Weather forecast in GeoJSON format */,\n \"annotations\": {\n \"audience\": [\"assistant\"]\n }\n }\n ]\n \"id\": 1\n}\n```\n\n## 디버깅\n\n[MCP Inspector는](https://github.com/modelcontextprotocol/inspector) MCP 서버를 테스트하고 디버깅하는 도구입니다. Claude 및/또는 emcee가 예상대로 작동하지 않을 경우, Inspector를 통해 문제를 파악할 수 있습니다.\n\n```console\nnpx @modelcontextprotocol/inspector emcee https://api.weather.gov/openapi.json\n# 🔍 MCP Inspector is up and running at http://localhost:5173 🚀\n```\n\n```console\nopen http://localhost:5173\n```\n\n## 특허\n\nemcee는 Apache 라이선스 버전 2.0에 따라 라이선스가 부여되었습니다.","Korean","ko-KR",{"_41":74,"_45":78,"_50":79},{"_43":75,"_45":76,"_47":77},"为任何 OpenAPI 记录的端点生成 MCP 服务器。","主持人","![主持人流程图](https://github.com/user-attachments/assets/bcac98a5-497f-4b34-9e8d-d4bc08852ea1)\n\n# 主持人\n\n**emcee**是一款工具，可为任何符合[OpenAPI](https://openapi.org)规范的 Web 应用程序提供[模型上下文协议 (MCP)](https://modelcontextprotocol.io/)服务器。您可以使用 emcee 将[Claude Desktop](https://claude.ai/download)和[其他应用程序](https://modelcontextprotocol.info/docs/clients/)连接到外部工具和数据服务，类似于[ChatGPT 插件](https://openai.com/index/chatgpt-plugins/)。\n\n## 快速入门\n\n如果您使用的是 macOS 并且安装了[Homebrew](https://brew.sh) ，则可以快速启动并运行。\n\n```bash\n# Install emcee\nbrew install loopwork-ai/tap/emcee\n```\n\n确保您已安装[Claude Desktop](https://claude.ai/download) 。\n\n要配置 Claude Desktop 以便与 emcee 一起使用：\n\n1. 打开 Claude 桌面设置 (⌘,)\n2. 选择侧边栏中的“开发者”部分\n3. 点击“编辑配置”打开配置文件\n\n![Claude 桌面设置编辑配置按钮](https://github.com/user-attachments/assets/761c6de5-62c2-4c53-83e6-54362040acd5)\n\n配置文件应该位于 Application Support 目录中。您也可以使用以下命令直接在 VSCode 中打开它：\n\n```console\ncode ~/Library/Application\\ Support/Claude/claude_desktop_config.json\n```\n\n添加以下配置以添加 weather.gov MCP 服务器：\n\n```json\n{\n \"mcpServers\": {\n \"weather\": {\n \"command\": \"emcee\",\n \"args\": [\n \"https://api.weather.gov/openapi.json\"\n ]\n }\n }\n}\n```\n\n保存文件后，退出并重新打开 Claude。您现在应该会在聊天框的右下角看到🔨57。点击它即可查看 Claude 通过 MCP 可以使用的所有工具列表。\n\n开始新的聊天并询问您所在地的天气。\n\n> 俄勒冈州波特兰的天气怎么样？\n\nClaude 将参考 MCP 提供的工具，如果其中一个工具被认为适合解答您的问题，Claude 将申请使用。您可以审核此请求，并决定批准或拒绝。\n\n如果您允许，Claude 将与 MCP 进行通信并使用结果来通知其响应。\n\n![Claude 对 MCP 工具使用的回应](https://github.com/user-attachments/assets/d5b63002-1ed3-4b03-bc71-8357427bb06b)\n\n***\n\n> 提示：正在构建代理？想要部署远程 MCP 服务器？请联系\n\n***\n\n## 为什么要使用 emcee？\n\nMCP 提供了一种将 AI 模型连接到工具和数据源的标准化方法。虽然目前尚处于早期阶段，但目前已有多种[可用于连接浏览器、开发者工具和其他系统的服务器](https://modelcontextprotocol.io/examples)。\n\n我们认为 emcee 是一种便捷的方式，可以连接到尚未实现 MCP 服务器的服务，*尤其适用于您自行构建的服务*。您有一个符合 OpenAPI 规范的 Web 应用吗？您可能会惊讶地发现，即使没有仪表板或客户端库，您也能取得如此大的进展。\n\n## 安装\n\n### 安装程序脚本\n\n使用[安装程序脚本](https://github.com/loopwork-ai/emcee/blob/main/tools/install.sh)下载并安装适合您平台（Linux x86-64/i386/arm64 和 macOS Intel/Apple Silicon）的[预构建版本](https://github.com/loopwork-ai/emcee/releases)emcee。\n\n```console\n# fish\nsh (curl -fsSL https://get.emcee.sh | psub)\n\n# bash, zsh\nsh <(curl -fsSL https://get.emcee.sh)\n```\n\n### 自制\n\n使用[Loopwork 的 tap](https://github.com/loopwork-ai/homebrew-tap)中的[Homebrew](https://brew.sh)安装 emcee。\n\n```console\nbrew install loopwork-ai/tap/emcee\n```\n\n### Docker\n\n带有 emcee 的预构建[Docker 镜像](https://github.com/loopwork-ai/emcee/pkgs/container/emcee)可供使用。\n\n```console\ndocker run -it ghcr.io/loopwork-ai/emcee\n```\n\n### 从源代码构建\n\n需要[1.24](https://go.dev)或更高版本。\n\n```console\ngit clone https://github.com/loopwork-ai/emcee.git\ncd emcee\ngo build -o emcee cmd/emcee/main.go\n```\n\n一旦构建完成，您就可以就地运行（ `./emcee` ）或将其移动到`PATH`中的某个位置，例如`/usr/local/bin` 。\n\n## 用法\n\n```console\nUsage:\n emcee [spec-path-or-url] [flags]\n\nFlags:\n --basic-auth string Basic auth value (either user:pass or base64 encoded, will be prefixed with 'Basic ')\n --bearer-auth string Bearer token value (will be prefixed with 'Bearer ')\n -h, --help help for emcee\n --raw-auth string Raw value for Authorization header\n --retries int Maximum number of retries for failed requests (default 3)\n -r, --rps int Maximum requests per second (0 for no limit)\n -s, --silent Disable all logging\n --timeout duration HTTP request timeout (default 1m0s)\n -v, --verbose Enable debug level logging to stderr\n --version version for emcee\n```\n\nemcee 为 MCP 实现[标准输入/输出 (stdio)](https://modelcontextprotocol.io/docs/concepts/transports#standard-input-output-stdio)传输，使用[JSON-RPC 2.0](https://www.jsonrpc.org/)作为其传输格式。\n\n当您从命令行运行 emcee 时，它会启动一个程序，该程序监听 stdin、输出到 stdout 并记录到 stderr。\n\n### 验证\n\n对于需要身份验证的API，emcee支持多种身份验证方法：\n\n| 身份验证类型 | 示例用法 | 结果标题 |\n| ---------- | ---------------------------- | ----------------------------------- |\n| **持有者令牌** | `--bearer-auth=\"abc123\"` | `Authorization: Bearer abc123` |\n| **基本身份验证** | `--basic-auth=\"user:pass\"` | `Authorization: Basic dXNlcjpwYXNz` |\n| **原始值** | `--raw-auth=\"Custom xyz789\"` | `Authorization: Custom xyz789` |\n\n这些身份验证值可以直接提供，也可以作为[1Password 秘密参考提供](https://developer.1password.com/docs/cli/secret-reference-syntax/)。\n\n使用 1Password 参考时：\n\n* 使用格式`op://vault/item/field` （例如`--bearer-auth=\"op://Shared/X/credential\"` ）\n* 确保 1Password CLI（ [op](https://developer.1password.com/docs/cli/get-started/) ）已安装并在您的`PATH`中可用\n* 在运行 emcee 或启动 Claude Desktop 之前登录 1Password\n\n```console\n# Install op\nbrew install 1password-cli\n\n# Sign in 1Password CLI\nop signin\n```\n\n```json\n{\n \"mcpServers\": {\n \"twitter\": {\n \"command\": \"emcee\",\n \"args\": [\n \"--bearer-auth=op://shared/x/credential\",\n \"https://api.twitter.com/2/openapi.json\"\n ]\n }\n }\n}\n```\n\n> \\[！重要的]\\\n> emcee 从命令参数提供的 URL 下载 OpenAPI 规范时不使用身份验证凭据。如果您的 OpenAPI 规范需要身份验证才能访问，请先使用您首选的 HTTP 客户端将其下载到本地文件，然后将本地文件路径提供给 emcee。\n\n### 转型 OpenAPI 规范\n\n您可以使用标准 Unix 实用程序转换 OpenAPI 规范，然后再将其传递给 emcee。这适用于：\n\n* 使用[jq](https://github.com/jqlang/jq)或[yq](https://github.com/mikefarah/yq)选择要公开为工具的特定端点\n* 使用[OpenAPI Overlays](https://www.openapis.org/blog/2024/10/22/announcing-overlay-specification)修改描述或参数\n* 使用[Redocly](https://redocly.com/docs/cli/commands)组合多种规格\n\n例如，您可以使用`jq`仅包含来自`weather.gov`的`point`工具。\n\n```console\ncat path/to/openapi.json | \\\n jq 'if .paths then .paths |= with_entries(select(.key == \"/points/{point}\")) else . end' | \\\n emcee\n```\n\n### JSON-RPC\n\n您可以通过发送 JSON-RPC 请求直接与提供的 MCP 服务器进行交互。\n\n> emcee 仅提供 MCP 工具功能。其他功能（例如资源、提示和采样）尚不受支持。\n\n#### 列表工具\n\n```json\n{\"jsonrpc\": \"2.0\", \"method\": \"tools/list\", \"params\": {}, \"id\": 1}\n```\n\n```jsonc\n{\n \"jsonrpc\":\"2.0\",\n \"result\": {\n \"tools\": [\n // ...\n {\n \"name\": \"tafs\",\n \"description\": \"Returns Terminal Aerodrome Forecasts for the specified airport station.\",\n \"inputSchema\": {\n \"type\": \"object\",\n \"properties\": {\n \"stationId\": {\n \"description\": \"Observation station ID\",\n \"type\": \"string\"\n }\n },\n \"required\": [\"stationId\"]\n }\n },\n // ...\n ]\n },\n \"id\": 1\n}\n```\n\n#### 呼叫工具\n\n```json\n{\"jsonrpc\": \"2.0\", \"method\": \"tools/call\", \"params\": { \"name\": \"taf\", \"arguments\": { \"stationId\": \"KPDX\" }}, \"id\": 1}\n```\n\n```jsonc\n{\n \"jsonrpc\":\"2.0\",\n \"content\": [\n {\n \"type\": \"text\",\n \"text\": /* Weather forecast in GeoJSON format */,\n \"annotations\": {\n \"audience\": [\"assistant\"]\n }\n }\n ]\n \"id\": 1\n}\n```\n\n## 调试\n\n[MCP 检查器](https://github.com/modelcontextprotocol/inspector)是一款用于测试和调试 MCP 服务器的工具。如果 Claude 和/或 emcee 未按预期工作，检查器可以帮助您了解问题所在。\n\n```console\nnpx @modelcontextprotocol/inspector emcee https://api.weather.gov/openapi.json\n# 🔍 MCP Inspector is up and running at http://localhost:5173 🚀\n```\n\n```console\nopen http://localhost:5173\n```\n\n## 执照\n\nemcee 已获得 Apache 许可证 2.0 版的许可。","Chinese","zh-CN","githubUser","locale",{"_83":84,"_85":86,"_28":99,"_102":44,"_103":44,"_104":105,"_108":109,"_131":-5,"_132":133,"_134":135,"_136":137,"_45":46,"_138":122,"_139":-5,"_140":-5,"_141":-5,"_142":-5,"_143":144,"_149":-5,"_150":118,"_151":152,"_155":-5,"_8":156,"_157":158},"addedAt",["D",1744308084163],"argumentsJsonSchema",{"_87":88,"_96":97,"_12":98},"properties",{"_89":90,"_93":94},"PUBMED_EMAIL",{"_43":91,"_12":92},"Your email address (required by NCBI)","string","PUBMED_API_KEY",{"_43":95,"_12":92},"Optional API key for higher rate limits","required",[89],"object",[100,101],"language:go","license_grade:a","descriptionMarkdown","descriptionPlainText","githubProject",{"_47":48,"_106":107},"url","https://github.com/loopwork-ai/emcee","githubRepository",{"_110":111,"_112":113,"_114":115,"_116":117,"_45":118,"_119":120,"_124":125,"_126":127,"_129":130},"defaultBranch","main","fullName","loopwork-ai/emcee","githubTreeId",106671,"language","Go","emcee","owner",{"_121":122,"_8":123},"login","loopwork-ai","JkY5JS2wPW","recentStargazers",9,"spdxLicense",{"_45":128},"Apache 2.0","stargazers",201,"glamaJson","id",5284,"inspectable",false,"integrations",[],"namespace","npmPackage","prompts","redditPostForeignId","resources","scores",{"_145":146,"_147":-5,"_148":-5},"license",100,"quality","security","simpleIcon","slug","supportedPlatforms",[153,154],"MACOS","LINUX","tools","hp6zmct9vh","updatedAt",["D",1744308329491],"meta",[161,163,165,172,175,177,180,183,186,189,192,194,197,200,202,204,206,208,210],{"_10":162},"Emcee | Glama",{"_164":44,"_45":43},"content",{"_166":167,"_168":169,"_170":171},"href","https://glama.ai/mcp/servers/@loopwork-ai/emcee","rel","canonical","tagName","link",{"_164":46,"_173":174},"property","og:title",{"_164":44,"_173":176},"og:description",{"_164":178,"_173":179},"https://glama.ai/logo.png","og:logo",{"_164":181,"_173":182},"https://glama.ai/mcp/servers/@loopwork-ai/emcee/og-image","og:image",{"_164":184,"_173":185},"630","og:image:height",{"_164":187,"_173":188},"1200","og:image:width",{"_164":190,"_173":191},"website","og:type",{"_164":167,"_173":193},"og:url",{"_164":195,"_173":196},"ChatGPT alternative for power users","og:site_name",{"_164":198,"_45":199},"summary_large_image","twitter:card",{"_164":195,"_45":201},"twitter:title",{"_164":44,"_45":203},"twitter:description",{"_164":181,"_45":205},"twitter:image",{"_164":184,"_45":207},"twitter:image:height",{"_164":187,"_45":209},"twitter:image:width",{"_211":212},"script:ld+json",{"_213":214,"_215":216},"@context","http://schema.org","@graph",[217],{"_213":218,"_219":220,"_221":222},"https://schema.org","@type","BreadcrumbList","itemListElement",[223,230,234,238],{"_219":224,"_225":226,"_45":227,"_228":229},"ListItem","item","https://glama.ai/","Glama","position",1,{"_219":224,"_225":231,"_45":232,"_228":233},"https://glama.ai/mcp","MCP",2,{"_219":224,"_225":235,"_45":236,"_228":237},"https://glama.ai/mcp/servers","Servers",3,{"_219":224,"_225":167,"_45":46,"_228":239},4,"visitorIsMaintainer","routes/_public/mcp/servers/~namespace/~slug/_pages/reviews/_route",{"_80":-5,"_41":243,"_159":259,"_293":-5},{"_83":244,"_85":245,"_28":250,"_102":44,"_103":44,"_104":251,"_108":252,"_131":-5,"_132":133,"_134":135,"_136":255,"_45":46,"_138":122,"_139":-5,"_140":-5,"_141":-5,"_142":-5,"_143":256,"_149":-5,"_150":118,"_151":257,"_155":-5,"_8":156,"_157":258},["D",1744308084163],{"_87":246,"_96":249,"_12":98},{"_89":247,"_93":248},{"_43":91,"_12":92},{"_43":95,"_12":92},[89],[100,101],{"_47":48,"_106":107},{"_110":111,"_112":113,"_114":115,"_116":117,"_45":118,"_119":253,"_124":125,"_126":254,"_129":130},{"_121":122,"_8":123},{"_45":128},[],{"_145":146,"_147":-5,"_148":-5},[153,154],["D",1744308329491],[260,262,263,265,267,268,269,270,271,272,273,274,275,276,277,278,279,280,281],{"_10":261},"Reviews | Emcee | Glama",{"_164":44,"_45":43},{"_166":264,"_168":169,"_170":171},"https://glama.ai/mcp/servers/@loopwork-ai/emcee/reviews",{"_164":266,"_173":174},"Reviews | Emcee",{"_164":44,"_173":176},{"_164":178,"_173":179},{"_164":181,"_173":182},{"_164":184,"_173":185},{"_164":187,"_173":188},{"_164":190,"_173":191},{"_164":264,"_173":193},{"_164":195,"_173":196},{"_164":198,"_45":199},{"_164":195,"_45":201},{"_164":44,"_45":203},{"_164":181,"_45":205},{"_164":184,"_45":207},{"_164":187,"_45":209},{"_211":282},{"_213":214,"_215":283},[284],{"_213":218,"_219":220,"_221":285},[286,287,288,289,290],{"_219":224,"_225":226,"_45":227,"_228":229},{"_219":224,"_225":231,"_45":232,"_228":233},{"_219":224,"_225":235,"_45":236,"_228":237},{"_219":224,"_225":167,"_45":46,"_228":239},{"_219":224,"_225":264,"_45":291,"_228":292},"Reviews",5,"review","actionData","errors"]

Rate & Review