@loop_ouroboros/mcp-hub-lite 1.1.1 → 1.2.1

package/CHANGELOG.md

@@ -1,5 +1,145 @@
 # Changelog
 
+## 2026-04-14
+
+- **Lint Fix**: Remove unnecessary escape character in CLI entry detection
+- **Docs**: Format README files for release
+- **Version**: Bump version to 1.2.1
+- **CLI Fix**: Fix CLI entry detection for Windows/npm symlink environments
+- **CLI Refactor**: Fix command count and rename mcpToolUseCommand to toolUseCommand
+- **Config Refactor**: Change config loading logs from info to debug level
+- **Core Feature**: Add normalizeToolName for cross-format tool name matching
+- **CLI Feature**: Add tool-use command for MCP server tool operations
+- **Gateway Feature**: Implement per-request transport mode to fix connection errors
+
+## 2026-04-13
+
+- **Deps**: Add undici dependency for HTTP transport
+- **Frontend Fix**: Prevent runtime error when tool is undefined in groupedTools
+- **Version**: Bump version to 1.1.0
+- **Core Refactor**: Use composite key serverName-serverIndex instead of serverId
+- **Resource Feature**: Implement MCP native resource forwarding
+- **Docs**: Update search functionality description in CLAUDE.md
+- **Hub Tools Refactor**: Add non-strict mode to selectBestInstance
+- **Search Refactor**: Remove complex search module, simplify to string matching
+- **Test Refactor**: Remove evaluation tests not compliant with MCP spec
+
+## 2026-04-12
+
+- **Config Refactor**: Move instance selection strategy to template
+- **Frontend Refactor**: Improve type declarations and test type safety
+- **Frontend Refactor**: Compute instance counts from server.instances directly
+- **Gateway Feature**: Restrict call_tool from invoking system tools
+- **Docs**: Compress use-guide from 13 to 7 sections
+- **Gateway Fix**: Return 400 for missing Accept header
+
+## 2026-04-10
+
+- **Transport Refactor**: Use official SDK StdioClientTransport
+- **Stdio Transport Refactor**: Log non-JSON-RPC stdout and unify stderr log level
+- **Frontend Fix**: Fix tools and resources not displaying in server detail
+- **Frontend Fix**: Add description field to ConfigTemplateForm
+- **Tools Fix**: Handle nested toolArgs in updateServerDescription
+
+## 2026-04-09
+
+- **Security Feature**: Mask sensitive values in config change logs
+- **Frontend Fix**: Improve instance status reactivity and management
+- **i18n Fix**: Improve noToolsFound message to reflect tool aggregation
+- **Frontend Fix**: Add bottom padding to prevent save button cutoff on scroll
+- **Gateway Refactor**: Extract version and protocol version to shared utility
+- **Config Refactor**: Unify instance ID generation with content hash
+
+## 2026-04-08
+
+- **MCP Status Fix**: Prevent automatic enabled field updates on connect/disconnect
+- **Chore**: Update .gitignore to ignore .full-review directory
+
+## 2026-04-07
+
+- **Frontend Fix**: Fix TypeScript type errors in component event handlers
+
+## 2026-04-04
+
+- **Test Refactor**: Simplify ServerStatusTags tests with mocks
+- **Config Feature**: Sort env and headers keys on config save
+- **Proxy Feature**: Add proxy configuration support for SSE and Streamable HTTP transports
+
+## 2026-04-03
+
+- **Config Feature**: Sort env and headers keys on config save
+- **Config Refactor**: Optimize updateServerInstance logging
+- **Connection Refactor**: Improve serverId handling and protocol version support
+
+## 2026-04-02
+
+- **Docs**: Update CLAUDE.md documentation for logging and composables
+- **Logging Refactor**: Enhance transport logging with serverId context
+- **Transports Feature**: Add LineBuffer for proper stderr line buffering
+- **Logging Refactor**: Enhance logging module and configuration change tracking
+- **Docs**: Update CLAUDE.md documentation for v1.1 config refactor
+
+## 2026-04-01
+
+- **Config Refactor**: Complete v1.1 multi-instance configuration refactoring
+- **Docs**: Update CLAUDE.md files to reflect session removal
+- **Session Refactor**: Remove session persistence mechanism and use SDK native stateless mode
+- **Session Refactor**: Remove session persistence mechanism
+
+## 2026-03-31
+
+- **UI Fix**: Ensure log viewer area fills available container height
+- **Frontend Feature**: Add instance tab routing synchronization
+
+## 2026-03-29
+
+- **i18n Fix**: Correct missing translation key in InstanceConfig component
+
+## 2026-03-28
+
+- **Hub Tools Feature**: Implement instance selection strategies
+
+## 2026-03-27
+
+- **Docs**: Update CLAUDE.md files across project modules
+- **Session Feature**: Remove cwd and project fields from session management
+- **UI Fix**: Optimize instance list display in English and hide redundant Running tag
+- **Config Fix**: Ensure server instance indexes from config load and migration
+
+## 2026-03-26
+
+- **Config Feature**: Implement v1.1 instance configuration system
+
+## 2026-03-25
+
+- **Config Fix**: Support string index type for server instance operations
+- **Docs**: Update CLAUDE.md files to reflect recent v1.1 config changes
+- **Config Feature**: Enhance instance management with display name and route selection
+
+## 2026-03-24
+
+- **Config Refactor**: Activate v1.1 config format with instance support
+- **Gateway Fix**: Only include tools from servers with allowedTools configured
+- **Hub Tools Fix**: Support both toolArgs and arguments for backward compatibility
+
+## 2026-03-23
+
+- **Config Refactor**: Activate v1.1 config format, remove v1.0 compatibility code
+- **Config Feature**: Add empty value cleaning when saving config
+
+## 2026-03-20
+
+- **UI Fix**: Correct dark mode background color and refine instance UI
+- **UI Feature**: Add server instance management UI components
+- **Config Feature**: Add index and displayName to ServerInstanceConfig with reassignment API
+- **UI Feature**: Add tag management UI with TagManager component
+- **Config Feature**: Add v1.1 config schema with auto-migration support
+
+## 2026-03-19
+
+- **i18n Fix**: Fix headers label and add button text in server config
+- **UI Feature**: Add description field support in AddServerModal
+
 ## 2026-03-17
 
 - **UI Enhancement**: Update resource detail and list views with hub tools integration

package/README.md

@@ -1,12 +1,12 @@
 # MCP-HUB-LITE
 
-[![License][license-src]][license-href]
-[![Node.js][node-src]][node-href]
-[![TypeScript][typescript-src]][typescript-href]
-[![Vitest][vitest-src]][vitest-href]
-[![Fastify][fastify-src]][fastify-href]
-[![Vue.js][vue-src]][vue-href]
-[![Claude Code][claude-code-src]][claude-code-href]
+[![License](https://img.shields.io/badge/license-MIT-080f12?style=flat&colorA=080f12&colorB=1fa669)](./LICENSE)
+[![Node.js](https://img.shields.io/badge/Node.js-22.x-080f12?style=flat&logo=nodedotjs&logoColor=white&colorA=080f12&colorB=339933)](https://nodejs.org/)
+[![TypeScript](https://img.shields.io/badge/TypeScript-5.x-080f12?style=flat&logo=typescript&logoColor=white&colorA=080f12&colorB=3178C6)](https://www.typescriptlang.org/)
+[![Vitest](https://img.shields.io/badge/Vitest-Testing-080f12?style=flat&logo=vitest&logoColor=white&colorA=080f12&colorB=6E9F18)](https://vitest.dev/)
+[![Fastify](https://img.shields.io/badge/Fastify-Web-080f12?style=flat&logo=fastify&logoColor=white&colorA=080f12&colorB=000000)](https://fastify.io/)
+[![Vue.js](https://img.shields.io/badge/Vue.js-3-080f12?style=flat&logo=vuedotjs&logoColor=white&colorA=080f12&colorB=4FC08D)](https://vuejs.org/)
+[![Claude Code](https://img.shields.io/badge/Claude-Code-1fa669?style=flat&colorA=080f12&colorB=1fa669)](https://claude.ai/code)
 
 ---
 
@@ -18,32 +18,22 @@ A lightweight MCP management platform designed for independent developers, provi
 
 MCP-HUB-LITE is an MCP server gateway designed specifically for independent developers. It acts as a proxy between your frontend and multiple backend MCP servers, providing a unified access interface with support for the MCP JSON-RPC 2.0 protocol.
 
-### Dashboard
-
-![Dashboard](./docs/screenshot/en/dashboard.png)
-
 ### Core Features
 
 - **MCP Gateway Service**: Unified proxy interface for multiple backend MCP servers
+- **MCP CLI Tool**: Command-line tool for listing and calling MCP tools across servers
 - **Server Management**: Manage multiple MCP servers through a web interface
-- **Tool Search**: Fuzzy search and tool discovery across all servers
+- **Tool Search**: Search and tool discovery across all servers with aggregation
 - **Process Management**: Launch and manage MCP server processes via npx/uvx
-- **Session Management**: Session state management with persistence and recovery
+- **Session Management**: Native stateless session management via MCP SDK
+- **Multi-Instance Support**: Run multiple instances of the same MCP server with load balancing
+- **Instance Selection Strategies**: Support random, round-robin, and unique-by-tag selection
 - **Tag System**: Organize multiple MCP servers by environment, category, function, etc.
 - **Fault Tolerance**: System continues to operate when individual servers fail
 - **Bilingual Interface**: Support for Chinese/English interface switching
 - **Configuration Management**: Support for hot-reloading and maintenance of `.mcp-hub.json`
-
-## Tech Stack
-
-- **TypeScript 5.x** + Node.js 22.x
-- **Fastify**: High-performance HTTP server
-- **MCP SDK**: Official MCP protocol support (@modelcontextprotocol/sdk)
-- **Vitest**: Unit testing framework
-- **Zod**: Data validation
-- **Vue 3**: Frontend UI framework
-- **Pinia**: Frontend state management
-- **Element Plus**: UI component library
+- **MCP Native Resources**: Forward resource calls to backend MCP servers
+- **Security**: Mask sensitive values in config change logs
 
 ## Quick Start
 
@@ -93,7 +83,7 @@ npm run status
 npm run ui
 ```
 
-The server will start at http://localhost:7788.
+The server will start at <http://localhost:7788>.
 
 ## Server Management
 
@@ -115,32 +105,17 @@ Discover and call tools from all connected MCP servers through the unified gatew
 
 Browse and manage MCP resources from all connected servers.
 
-## Sessions
-
-![Sessions](./docs/screenshot/en/sessions.png)
-
-Session management with persistence support. Sessions are automatically saved to disk and can be restored after service restart.
-
-## Settings
-
-![Settings](./docs/screenshot/en/settings.png)
-
-Configure language, logging, and other settings through the settings page.
-
 ### Testing
 
 ```bash
 # Run all tests
 npm test
 
-# Unit tests
-npm run test:unit
-
-# Integration tests
-npm run test:integration
+# Backend tests
+npm run test:backend
 
-# Contract tests
-npm run test:contract
+# Frontend tests
+npm run test:frontend
 ```
 
 ## CLI Commands
@@ -166,6 +141,32 @@ node dist/index.js ui
 node dist/index.js --help
 ```
 
+### Tool Use Command
+
+The `tool-use` command provides MCP server tool operations:
+
+```bash
+# List system tools (default server: mcp-hub-lite)
+npm run tool-use -- list-tools
+mcp-hub-lite tool-use list-tools
+
+# List tools from a specific server
+npm run tool-use -- list-tools --server baidu-search
+mcp-hub-lite tool-use list-tools --server baidu-search
+
+# Get tool schema
+npm run tool-use -- get-tool --tool list_tools
+mcp-hub-lite tool-use get-tool --tool list_tools
+
+# Call a system tool
+npm run tool-use -- call-tool --tool list_tools --args '{}'
+mcp-hub-lite tool-use call-tool --tool list_tools --args '{}'
+
+# Call a server tool
+npm run tool-use -- call-tool --server baidu-search --tool search --args '{"query":"hello"}'
+mcp-hub-lite tool-use call-tool --server baidu-search --tool search --args '{"query":"hello"}'
+```
+
 ## Configuration
 
 MCP-HUB-LITE uses a `.mcp-hub.json` file for configuration. Configuration lookup priority:
@@ -177,24 +178,33 @@ MCP-HUB-LITE uses a `.mcp-hub.json` file for configuration. Configuration lookup
 
 ```json
 {
-  "version": "1.0.0",
+  "version": "1.1.0",
   "servers": [
     {
       "id": "server-1",
       "name": "My MCP Server",
       "description": "Example server",
+      "transport": "streamable-http",
       "endpoint": "http://localhost:8080",
-      "transport": "http-stream",
       "tags": {
         "env": "development",
         "category": "api-server",
         "function": "http-api",
         "priority": "medium"
       },
+      "allowedTools": [],
+      "instances": [
+        {
+          "index": 0,
+          "displayName": "Instance 1",
+          "enabled": true,
+          "env": {}
+        }
+      ],
       "managedProcess": {
         "command": "npx my-mcp-server",
         "managedMode": "npx",
-        "processType": "http-stream"
+        "processType": "streamable-http"
       }
     }
   ],
@@ -225,50 +235,11 @@ MCP-HUB-LITE uses a `.mcp-hub.json` file for configuration. Configuration lookup
 
 Through the web interface:
 
-1. Open http://localhost:7788
+1. Open <http://localhost:7788>
 2. Navigate to the "Servers" page
 3. Click "Add Server"
 4. Fill in server details and save
 
-### MCP Protocol Usage
-
-MCP-HUB-LITE exposes the MCP protocol interface on the same port:
-
-#### List All Tools
-
-```json
-{
-  "jsonrpc": "2.0",
-  "method": "tools/list",
-  "params": {}
-}
-```
-
-#### Call a Tool
-
-```json
-{
-  "jsonrpc": "2.0",
-  "method": "tool/call",
-  "params": {
-    "toolId": "server1:example-tool",
-    "arguments": {}
-  }
-}
-```
-
-#### Get Server Status
-
-```json
-{
-  "jsonrpc": "2.0",
-  "method": "server/status",
-  "params": {
-    "serverId": "server-1"
-  }
-}
-```
-
 ## Process Management
 
 MCP-HUB-LITE supports launching and managing MCP servers using your local environment:
@@ -297,9 +268,15 @@ src/
 │   └── web-api/      # Web API routes
 ├── models/           # Data models
 ├── services/         # Core business logic
+│   ├── gateway/      # MCP gateway service
+│   ├── connection/   # Connection management
+│   └── hub-tools/    # Hub tools service
 ├── utils/            # Utility functions
+│   ├── logger/       # Logging utilities
+│   └── transports/   # MCP transport implementations
 ├── config/           # Configuration
 ├── cli/              # CLI commands
+│   └── commands/     # CLI command implementations
 ├── pid/              # Process ID management
 └── server/           # Server runtime
 
@@ -308,17 +285,21 @@ frontend/
 │   ├── components/   # Reusable UI components
 │   ├── views/        # Page view components
 │   ├── stores/       # Pinia state management
+│   ├── composables/  # Vue composables
 │   ├── router/       # Vue Router configuration
-│   └── i18n/         # Internationalization
+│   ├── i18n/         # Internationalization
+│   └── types/        # Frontend type definitions
 
 shared/
 ├── models/           # Shared models
-└── types/            # Shared types
+└── types/           # Shared types
 
 tests/
-├── unit/             # Unit tests
-├── integration/      # Integration tests
-└── contract/         # Contract tests
+├── unit/            # Unit tests
+├── integration/     # Integration tests
+├── contract/        # Contract tests
+├── helpers/         # Test helpers
+└── types/          # Test types
 ```
 
 ### Adding New Features
@@ -329,14 +310,6 @@ tests/
 4. Write tests (tests/)
 5. Update configuration files
 
-## Constraints and Limits
-
-- Maximum servers: 50
-- Maximum memory usage: 4GB
-- CPU usage threshold: 80%
-- Search response time: <500ms (90%)
-- Gateway latency: <100ms
-
 ## Detailed Technical Documentation
 
 Complete project architecture, constraints, and design decisions can be found in:
@@ -352,18 +325,3 @@ MIT
 Pull Requests and Issues are welcome!
 
 <!-- Badges -->
-
-[license-src]: https://img.shields.io/badge/license-MIT-080f12?style=flat&colorA=080f12&colorB=1fa669
-[license-href]: ./LICENSE
-[node-src]: https://img.shields.io/badge/Node.js-22.x-080f12?style=flat&logo=nodedotjs&logoColor=white&colorA=080f12&colorB=339933
-[node-href]: https://nodejs.org/
-[typescript-src]: https://img.shields.io/badge/TypeScript-5.x-080f12?style=flat&logo=typescript&logoColor=white&colorA=080f12&colorB=3178C6
-[typescript-href]: https://www.typescriptlang.org/
-[vitest-src]: https://img.shields.io/badge/Vitest-Testing-080f12?style=flat&logo=vitest&logoColor=white&colorA=080f12&colorB=6E9F18
-[vitest-href]: https://vitest.dev/
-[fastify-src]: https://img.shields.io/badge/Fastify-Web-080f12?style=flat&logo=fastify&logoColor=white&colorA=080f12&colorB=000000
-[fastify-href]: https://fastify.io/
-[vue-src]: https://img.shields.io/badge/Vue.js-3-080f12?style=flat&logo=vuedotjs&logoColor=white&colorA=080f12&colorB=4FC08D
-[vue-href]: https://vuejs.org/
-[claude-code-src]: https://img.shields.io/badge/Claude-Code-1fa669?style=flat&colorA=080f12&colorB=1fa669
-[claude-code-href]: https://claude.ai/code

package/dist/server/src/cli/commands/mcp-tool-use.d.ts

@@ -0,0 +1,62 @@
+import { Command } from 'commander';
+/**
+ * CLI command for dynamic MCP server tool operations via API.
+ *
+ * This command provides a simplified CLI interface for interacting with MCP server tools,
+ * supporting four actions: list-servers, list-tools, get-tool, and call-tool. It wraps
+ * the HTTP API endpoints and requires the MCP Hub Lite server to be running.
+ *
+ * ## Command Format
+ *
+ * ```
+ * mcp-tool-use <serverName> <action> [toolName] [--args <json>] [--tags <json>]
+ * ```
+ *
+ * Or via npm:
+ * ```
+ * npm run tool-use <serverName> <action> [toolName] [--args <json>] [--tags <json>]
+ * ```
+ *
+ * ## Supported Actions
+ *
+ * - `list-servers` - List all connected MCP servers
+ * - `list-tools` - List all tools from the specified server
+ * - `get-tool <toolName>` - Get complete schema for a specific tool
+ * - `call-tool <toolName>` - Call a tool on the specified server
+ *
+ * ## Options
+ *
+ * - `--args <json>` - JSON string of tool arguments (for call-tool action)
+ * - `--tags <json>` - JSON object of instance selection tags (call-tool only, for multi-instance servers)
+ *
+ * ## Usage Examples
+ *
+ * ```bash
+ * # List all connected servers
+ * mcp-hub-lite mcp-tool-use mcp-hub-lite list-servers
+ *
+ * # List all tools from a specific server
+ * mcp-hub-lite mcp-tool-use mcp-hub-lite list-tools
+ *
+ * # Get tool schema
+ * mcp-hub-lite mcp-tool-use mcp-hub-lite get-tool list_tools
+ *
+ * # Call a tool with arguments
+ * mcp-hub-lite mcp-tool-use mcp-hub-lite call-tool list_tools --args '{"serverName":"mcp-hub-lite"}'
+ *
+ * # Call a tool with tags for multi-instance selection
+ * mcp-hub-lite mcp-tool-use baidu-search call-tool search --args '{"query":"天气"}' --tags '{"env":"prod"}'
+ * ```
+ *
+ * ## Error Handling
+ *
+ * - Exits with code 1 if the server is not running
+ * - Exits with code 1 if action is unknown
+ * - Exits with code 1 if toolName is required but not provided
+ * - Exits with code 1 if JSON parsing fails for --args or --tags
+ * - Exits with code 1 if the underlying API call fails
+ *
+ * @returns {Command} The configured mcp-tool-use command instance for registration with Commander.js
+ */
+export declare const mcpToolUseCommand: Command;
+//# sourceMappingURL=mcp-tool-use.d.ts.map

package/dist/server/src/cli/commands/mcp-tool-use.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"mcp-tool-use.d.ts","sourceRoot":"","sources":["../../../../../src/cli/commands/mcp-tool-use.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,OAAO,EAAE,MAAM,WAAW,CAAC;AAGpC;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA0DG;AACH,eAAO,MAAM,iBAAiB,SAkI1B,CAAC"}

package/dist/server/src/cli/commands/mcp-tool-use.js

@@ -0,0 +1,174 @@
+import { Command } from 'commander';
+import { getServerStatus } from '../server.js';
+/**
+ * CLI command for dynamic MCP server tool operations via API.
+ *
+ * This command provides a simplified CLI interface for interacting with MCP server tools,
+ * supporting four actions: list-servers, list-tools, get-tool, and call-tool. It wraps
+ * the HTTP API endpoints and requires the MCP Hub Lite server to be running.
+ *
+ * ## Command Format
+ *
+ * ```
+ * mcp-tool-use <serverName> <action> [toolName] [--args <json>] [--tags <json>]
+ * ```
+ *
+ * Or via npm:
+ * ```
+ * npm run tool-use <serverName> <action> [toolName] [--args <json>] [--tags <json>]
+ * ```
+ *
+ * ## Supported Actions
+ *
+ * - `list-servers` - List all connected MCP servers
+ * - `list-tools` - List all tools from the specified server
+ * - `get-tool <toolName>` - Get complete schema for a specific tool
+ * - `call-tool <toolName>` - Call a tool on the specified server
+ *
+ * ## Options
+ *
+ * - `--args <json>` - JSON string of tool arguments (for call-tool action)
+ * - `--tags <json>` - JSON object of instance selection tags (call-tool only, for multi-instance servers)
+ *
+ * ## Usage Examples
+ *
+ * ```bash
+ * # List all connected servers
+ * mcp-hub-lite mcp-tool-use mcp-hub-lite list-servers
+ *
+ * # List all tools from a specific server
+ * mcp-hub-lite mcp-tool-use mcp-hub-lite list-tools
+ *
+ * # Get tool schema
+ * mcp-hub-lite mcp-tool-use mcp-hub-lite get-tool list_tools
+ *
+ * # Call a tool with arguments
+ * mcp-hub-lite mcp-tool-use mcp-hub-lite call-tool list_tools --args '{"serverName":"mcp-hub-lite"}'
+ *
+ * # Call a tool with tags for multi-instance selection
+ * mcp-hub-lite mcp-tool-use baidu-search call-tool search --args '{"query":"天气"}' --tags '{"env":"prod"}'
+ * ```
+ *
+ * ## Error Handling
+ *
+ * - Exits with code 1 if the server is not running
+ * - Exits with code 1 if action is unknown
+ * - Exits with code 1 if toolName is required but not provided
+ * - Exits with code 1 if JSON parsing fails for --args or --tags
+ * - Exits with code 1 if the underlying API call fails
+ *
+ * @returns {Command} The configured mcp-tool-use command instance for registration with Commander.js
+ */
+export const mcpToolUseCommand = new Command('tool-use')
+    .description('Manage MCP server tools via API (list-servers, list-tools, get-tool, call-tool)')
+    .argument('<serverName>', 'Server name to target (use "mcp-hub-lite" for system tools)')
+    .argument('<action>', 'Action: list-servers, list-tools, get-tool, call-tool')
+    .argument('[toolName]', 'Tool name (required for get-tool and call-tool actions)')
+    .option('--args <json>', 'JSON string of tool arguments (call-tool only)')
+    .option('--tags <json>', 'JSON object of instance selection tags (call-tool only, for multi-instance servers)')
+    .action(async (serverName, action, toolName, options) => {
+    try {
+        // Check if server is running and get connection info
+        const status = await getServerStatus();
+        if (!status.running) {
+            console.error('Error: MCP Hub Lite server is not running.');
+            console.error('Start the server with: mcp-hub-lite start');
+            process.exit(1);
+        }
+        const baseUrl = `http://${status.host}:${status.port}`;
+        switch (action) {
+            case 'list-servers': {
+                const response = await fetch(`${baseUrl}/web/hub-tools/servers`, {
+                    headers: { Accept: 'application/json' }
+                });
+                if (!response.ok) {
+                    throw new Error(`API error: ${response.status} ${response.statusText}`);
+                }
+                const result = await response.json();
+                console.log(JSON.stringify(result, null, 2));
+                break;
+            }
+            case 'list-tools': {
+                const tagsParam = options.tags ? `?tags=${encodeURIComponent(options.tags)}` : '';
+                const response = await fetch(`${baseUrl}/web/hub-tools/servers/${serverName}/tools${tagsParam}`, {
+                    headers: { Accept: 'application/json' }
+                });
+                if (!response.ok) {
+                    const error = await response.json().catch(() => ({ message: response.statusText }));
+                    throw new Error(error.message || `API error: ${response.status}`);
+                }
+                const result = await response.json();
+                console.log(JSON.stringify(result, null, 2));
+                break;
+            }
+            case 'get-tool': {
+                if (!toolName) {
+                    console.error('Error: toolName is required for get-tool action');
+                    process.exit(1);
+                }
+                const tagsParam = options.tags ? `?tags=${encodeURIComponent(options.tags)}` : '';
+                const response = await fetch(`${baseUrl}/web/hub-tools/servers/${serverName}/tools/${toolName}${tagsParam}`, { headers: { Accept: 'application/json' } });
+                if (!response.ok) {
+                    const error = await response.json().catch(() => ({ message: response.statusText }));
+                    throw new Error(error.message || `API error: ${response.status}`);
+                }
+                const result = await response.json();
+                console.log(JSON.stringify(result, null, 2));
+                break;
+            }
+            case 'call-tool': {
+                if (!toolName) {
+                    console.error('Error: toolName is required for call-tool action');
+                    process.exit(1);
+                }
+                // Parse tool arguments
+                let toolArgs = {};
+                if (options.args) {
+                    try {
+                        toolArgs = JSON.parse(options.args);
+                    }
+                    catch {
+                        console.error('Error: Invalid JSON in --args option');
+                        process.exit(1);
+                    }
+                }
+                // Parse tags for instance selection (call-tool only)
+                let requestOptions;
+                if (options.tags) {
+                    try {
+                        requestOptions = { tags: JSON.parse(options.tags) };
+                    }
+                    catch {
+                        console.error('Error: Invalid JSON in --tags option');
+                        process.exit(1);
+                    }
+                }
+                const response = await fetch(`${baseUrl}/web/hub-tools/servers/${serverName}/tools/${toolName}/call`, {
+                    method: 'POST',
+                    headers: {
+                        'Content-Type': 'application/json',
+                        Accept: 'application/json'
+                    },
+                    body: JSON.stringify({ toolArgs, requestOptions })
+                });
+                if (!response.ok) {
+                    const error = await response.json().catch(() => ({ message: response.statusText }));
+                    throw new Error(error.message || `API error: ${response.status}`);
+                }
+                const result = await response.json();
+                console.log(JSON.stringify(result, null, 2));
+                break;
+            }
+            default: {
+                console.error(`Unknown action: ${action}`);
+                console.error('Valid actions: list-servers, list-tools, get-tool, call-tool');
+                process.exit(1);
+            }
+        }
+        process.exit(0);
+    }
+    catch (error) {
+        console.error(`Error: ${error instanceof Error ? error.message : String(error)}`);
+        process.exit(1);
+    }
+});