MCP solved how AI agents integrate with other systems. MCPB solved how users install them. But if you’re building MCP tools, you’re still copying JSON configs, wrestling with dependencies, and manually testing against clients.Documentation Index
Fetch the complete documentation index at: https://mintlify.com/superradcompany/tool-cli/llms.txt
Use this file to discover all available pages before exploring further.
tool-cli is the missing piece. It handles the entire lifecycle — from scaffolding to publishing — so you can focus on building your tool.
The MCPB ecosystem
Each layer in the ecosystem builds on the one below it:| Layer | Role |
|---|---|
| MCP protocol | Defines how AI agents communicate with tool servers |
| MCPB / MCPBX format | Defines how MCP tools are packaged and distributed |
| tool-cli toolchain | Scaffolds, tests, bundles, and publishes MCPB packages |
| tool.store registry | The discovery and distribution hub for published tools |
tool-cli is the toolchain layer. tool.store is the registry. Together they make building and sharing MCP tools as easy as publishing an npm package.
tool-cli also introduces MCPBX (
.mcpbx), a superset of MCPB that adds HTTP transport, reference mode (for npx/uvx or remote URLs), OAuth config, and template functions for auth headers — covering the cases the original spec doesn’t.Key capabilities
- Growing marketplace — discover and install MCP tools from tool.store
- Works with your stack — Claude Code, Cursor, OpenCode, VS Code, and more
- Ship your own MCP — create a new MCP server interactively with
tool init - Unified proxy — run all your MCPs through a single
tool runinterface - MCPs as CLIs — invoke any tool method directly from your terminal with
tool call - Configure once — set it up once, use it everywhere with
tool host - Built for humans and agents — clean output that works in terminals and AI workflows
- Encrypted by default — API keys and secrets are encrypted at rest
- OAuth just works — browser flow, token refresh, and secure storage are handled for you
- Fully local — your API keys and tokens never leave your machine
What problems it solves
Manual JSON configs. Every time you want to register an MCP server with a host like Claude Desktop or Cursor, you edit a JSON config by hand.tool host add handles this for all supported hosts.
No discovery. There was no central place to find and install MCP tools. tool search and tool.store fill that gap.
No testing workflow. Testing MCP servers required a full client setup. tool info shows what your server exposes and tool call lets you invoke any method directly — no client needed.
No build and publish pipeline. Bundling, cross-platform packaging, and publishing required custom scripts. tool pack and tool publish handle all of it, including multi-platform bundles for tools with native dependencies.
Next steps
Quick start
Install a tool, call it from the terminal, and register it with your AI host in minutes.
Command reference
Full reference for every tool-cli command.