build-mcpb

Package local MCP servers with bundled runtimes into installable .mcpb archives.

1|Updated Jul 6, 2015
One-click install
npx skills add https://github.com/ksolomon/dotfiles --skill build-mcpb-ksolomon
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: build-mcpb
Source: https://github.com/ksolomon/dotfiles/tree/main/AI/.claude/plugins/marketplaces/claude-plugins-official/plugins/mcp-server-dev/skills/build-mcpb
Command: npx skills add https://github.com/ksolomon/dotfiles --skill build-mcpb-ksolomon

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires @anthropic-ai/mcpb, @modelcontextprotocol/sdk, esbuild, zod, and includes references (resource) components.

What problem does it solve? Distributing a local MCP server normally requires users to install Node or Python and manage dependencies themselves. This Skill guides you through packaging your server with its runtime into a single .mcpb file that installs in Claude Desktop without any toolchain on the user's machine. ## Core Features & Use Cases - Manifest Authoring: Create a valid manifest.json (v0.4) with entry points, launch config, install-time user settings, and platform compatibility gates. - Build Pipeline: Bundle Node servers with esbuild or vendor Python dependencies, then validate and pack with the @anthropic-ai/mcpb CLI. - Local Security Guidance: Implement path containment checks, command injection prevention, read-only tool splits, and resource limits, since MCPB provides no sandbox. - Use Case: You built an MCP server that reads local files for users. Use this Skill to write the manifest, bundle dependencies, pack and sign the .mcpb, and ship a file users install by dragging it onto Claude Desktop. ## Quick Start Package my local MCP server into an installable .mcpb bundle with a proper manifest and bundled dependencies.

Frequently Asked Questions about build-mcpb

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I package an MCP server for distribution?▼

Bundle your server code and dependencies into a directory with a manifest.json, then run npx @anthropic-ai/mcpb pack to validate and zip it into a .mcpb file. Users install it by dragging the file onto Claude Desktop.

When should I use MCPB instead of a remote MCP server?▼

Use MCPB only when the server must run on the user's machine, such as reading local files, driving desktop apps, or calling OS APIs. If your server only hits cloud APIs, a remote HTTP server is the recommended distribution path.

Does MCPB provide a sandbox or permissions system?▼

No. MCPB has no permissions block and the server runs with full user privileges. You must enforce security in your tool handlers: validate path containment, avoid shell execution, split read and write tools, and cap resource usage.

How do I bundle Python dependencies in an MCPB package?▼

Install dependencies into a subdirectory with pip install -t server/vendor -r requirements.txt, prepend that directory to sys.path in your entry script, then run mcpb pack. Avoid native extensions since they must be built per platform.

Why does my MCPB server work on my machine but fail for users?▼

This almost always means a dependency was not actually bundled into the archive. Test the packed .mcpb on a machine without your dev toolchain installed before shipping.