MCP Setup Guide
Connecting PromptingBox to your AI tools takes about two minutes. The PromptingBox MCP server is published as an npm package (@promptingbox/mcp) and runs locally on your machine. Once connected, you can save prompts, search your library, manage folders, and retrieve any prompt -- all from inside Claude, Cursor, or any MCP-compatible tool without switching tabs.
To set up, first get your API key from the PromptingBox settings page. For the Claude desktop app, add a server entry to your claude_desktop_config.json file with the command "npx @promptingbox/mcp" and your API key as an environment variable. For Cursor, add the same config to .cursor/mcp.json in your project. For Claude Code (CLI), run "claude mcp add" with the server name "pbox". Each setup method takes one config block -- no build steps, no compilation, no Docker required.
Once connected, you can interact with your prompt library naturally. Say "save this as my code review prompt" and it saves. Say "find my API design prompt" and it searches by title and content. You can list folders, move prompts between them, browse version history, and even search community templates -- all through conversation. The MCP server handles authentication via your API key and keeps your prompts synced across every connected tool.
MCP Configuration Prompts
Prompts for setting up, testing, and managing MCP server connections.
MCP Server Config Generator
Generate an MCP server configuration block for {{ai_tool}} that connects to {{service_name}}. Server details: - Command: {{server_command}} - Environment variables: {{env_vars}} - Transport: stdio Output the exact JSON config block I should add to the appropriate config file. Include comments explaining each field. Also tell me which file to edit and where it lives on {{operating_system}}.
Why it works: Specifying the exact tool, service, and OS eliminates ambiguity and produces a paste-ready config block.
MCP Tool Definition Writer
Write an MCP tool definition for a function called "{{tool_name}}" that {{tool_description}}. Parameters: {{parameters_list}} Return format: {{return_format}} Generate the complete tool schema in JSON following the MCP specification, including: 1. name and description 2. inputSchema with JSON Schema types and required fields 3. Example invocation with sample parameters 4. Example successful response
Why it works: Providing the parameter list and return format upfront lets the model produce a complete, valid schema instead of a generic skeleton.
MCP Connection Test Script
Write a diagnostic script that tests whether my MCP server "{{server_name}}" is properly connected and responding. The script should: 1. Verify the server process is running 2. Send a basic initialization handshake 3. List available tools from the server 4. Call a simple test tool (like list or ping) and verify the response 5. Report success/failure with clear error messages for each step Target environment: {{runtime}} on {{operating_system}} Server command: {{server_command}}
Why it works: Step-by-step diagnostics isolate exactly where a connection fails instead of giving a generic 'not working' error.
MCP Function Schema Designer
I'm building an MCP server that exposes {{service_description}} to AI assistants. Design the complete function schema for these operations: {{operations_list}} For each operation, provide: 1. Function name (following MCP naming conventions) 2. Description (clear enough for an AI to decide when to call it) 3. Input parameters with types, descriptions, and required/optional flags 4. Response schema 5. Error cases and appropriate error codes Ensure the naming is consistent and the descriptions help the AI understand when to use each tool.
Why it works: Asking for descriptions 'clear enough for an AI to decide when to call it' produces schemas optimized for AI tool selection, not just human readability.
MCP Error Handler
My MCP server "{{server_name}}" is returning the following error: {{error_message}} Context: - AI tool: {{ai_tool}} - Server command: {{server_command}} - Config file location: {{config_path}} - Last working state: {{last_working_description}} Diagnose the issue and provide: 1. Most likely cause 2. Step-by-step fix 3. How to verify the fix worked 4. Preventive measures to avoid this error in the future
Why it works: Including the last working state helps narrow the diagnosis to what changed, dramatically reducing debugging time.
Multi-Server MCP Orchestrator
I need to configure {{ai_tool}} to connect to multiple MCP servers simultaneously: {{servers_list}} For each server, I have the command and API keys ready. Generate: 1. The complete config file with all servers defined 2. Naming conventions that avoid conflicts 3. A testing checklist to verify each server independently 4. Tips for handling cases where multiple servers expose similar tool names 5. Performance considerations for running {{server_count}} servers concurrently
Why it works: Addressing naming conflicts and concurrent performance upfront prevents the most common multi-server setup failures.
Recommended tools & resources
Understand the protocol behind AI tool integrations.
AI Tool ConfigsBrowse CLAUDE.md and .cursorrules configs for your projects.
Save Prompts Across ToolsWhy a central prompt library matters for multi-tool workflows.
Prompt TemplatesTemplates you can access via MCP from Claude and Cursor.
Model GuidesGet the most from Claude, Cursor, and other connected tools.
Prompt BuilderCreate new prompts and save them to your MCP-connected library.