MCP Config Builder - Free Online Tool | PivaBox

Visually build MCP server configurations for Claude Code — add servers, set commands, and export as JSON

MCP Config Builder — Visually Construct Model Context Protocol Server Configurations for Claude Code with Presets, Validation, and JSON Export

Add one or more MCP servers to your configuration by clicking Add Server. For each server, provide a unique server name (used as the configuration key), select the transport type — stdio for local command-line tools (the MCP server process is spawned as a child process by Claude Code) or SSE (Server-Sent Events) for connecting to remote MCP servers over HTTP. For stdio transport, specify the command (e.g., npx, node, python, or a direct binary path) and any arguments the command needs. For SSE transport, provide the server URL (e.g., http://localhost:3000/sse). Optionally add environment variables and custom headers for each server.
Use the Presets feature to quickly scaffold common MCP server configurations — select from popular community servers like filesystem, GitHub, PostgreSQL, Brave Search, Puppeteer, and more. Each preset comes pre-configured with the correct command, arguments, and common environment variables. After applying a preset, you can customize any field — change the server name, add extra arguments, or configure environment-specific variables. For fully custom servers, build from scratch — the form validates each field in real-time and highlights missing or invalid entries so you can fix issues before exporting.
Review the live JSON Preview panel showing your complete MCP configuration in real-time. The output follows the standard Claude Code .mcp.json format — an object where each key is a server name and each value contains the transport configuration. Click Copy JSON to grab the configuration to your clipboard, or Download to save as .mcp.json for placement in your project root directory. Claude Code automatically discovers this file and connects to the configured MCP servers, giving the AI assistant access to external tools, APIs, databases, and services defined in your configuration.

Frequently Asked Questions

What is the Model Context Protocol (MCP) and how does it extend Claude Code's capabilities?

The Model Context Protocol (MCP) is an open standard developed by Anthropic that enables AI assistants like Claude to interact with external tools, data sources, and services through a standardized interface. Think of MCP as 'USB-C for AI tools' — before MCP, every AI-tool integration was a custom one-off implementation. With MCP, any tool that implements the MCP server specification can be discovered and used by any MCP-compatible client (like Claude Code). An MCP server provides: Tools (functions the AI can call, like file operations or database queries), Resources (data the AI can read, like documentation or database schemas), and Prompts (pre-defined prompt templates). The PivaBox MCP Config Builder simplifies the process of creating the <code>.mcp.json</code> configuration file that tells Claude Code which MCP servers to connect to. All configuration is built locally in your browser — your server settings and environment variables never leave your device.

What's the difference between stdio and SSE transport, and when should I use each?

The two transport types serve different deployment scenarios. stdio (standard input/output) is the most common transport — Claude Code spawns the MCP server as a child process and communicates via standard input/output streams. This is ideal for local tools where the MCP server runs on the same machine: filesystem access, local databases, code linters, and any tool that can be launched from the command line. Stdio is simpler to set up (no network configuration needed), more secure (no open ports), and has lower latency (no network round-trips). SSE (Server-Sent Events) is designed for remote MCP servers — the server runs as a web service accessible via HTTP, and Claude Code connects to it using SSE for server-to-client messages and HTTP POST for client-to-server messages. Use SSE when: the MCP server runs on a different machine (team-shared development servers, cloud-hosted tools), the server needs to be shared across multiple Claude Code instances, or the server is implemented in a language or framework that doesn't work well as a CLI subprocess. The PivaBox builder supports both transports with fields that appear contextually based on your selection.

Where should I place the MCP configuration file, and can I have project-specific and global configurations?

MCP configurations support a hierarchy similar to <code>.gitignore</code> and <code>.eslintrc</code>. Project-level: place <code>.mcp.json</code> in your project root directory — Claude Code automatically discovers it when launched from that directory. Project-level configs are version-controlled (commit to git) so your entire team gets the same MCP server setup. Global-level: configure MCP servers globally via Claude Code's settings (<code>~/.claude/settings.json</code>) — these apply to all projects. Use global config for personal productivity tools (notes, calendar, email) and project config for team-shared development tools (package registry, deployment CLI, database tools). Claude Code merges both configurations, with project-level servers taking precedence in case of naming conflicts. The PivaBox MCP Config Builder helps you create either type — use it to prototype your configuration visually, then save the output to the appropriate location.