MCP server for discovering and serving Claude Code plugins
$git clone https://github.com/provectus/claude-pluginsInstalls into the current project.
Install claude-plugins by running `git clone https://github.com/provectus/claude-plugins`, then use it for the current task and follow its documentation at https://github.com/provectus/claude-plugins.
| 1 | # Claude Plugins |
| 2 | |
| 3 | A shared repository of plugins for Claude Code. Each plugin packages reusable expertise — agents, skills, prompts, and MCP server configs — into a standard format that can be discovered and consumed through the built-in MCP server. Available as an [npm package](https://www.npmjs.com/package/@provectusinc/claude-plugins) for easy integration via `npx`. |
| 4 | |
| 5 | ## Philosophy |
| 6 | |
| 7 | **Codify expertise, not just code.** The best engineering knowledge lives in people's heads — how to structure a Kotlin coroutine scope, when to reach for React Server Components, which Python async patterns actually work in production. This repo turns that knowledge into portable, versionable plugins that any Claude Code session can pull in on demand. |
| 8 | |
| 9 | A few principles guide the design: |
| 10 | |
| 11 | - **Convention over configuration.** Drop files in the right directories and the system picks them up. No central registry to maintain, no config files to wire together. |
| 12 | - **Composable pieces.** Plugins are made of independent components (agents, skills, prompts, MCP configs). Use what you need, ignore what you don't. |
| 13 | - **Plain text wins.** Plugin logic lives in Markdown files. They're easy to read, easy to review, and easy to edit without tooling. |
| 14 | - **Discovery is built in.** The repo itself is an MCP server. Other tools and Claude Code instances can search and retrieve plugins programmatically via `list_plugins`, `get_plugin`, and `search_plugins`. |
| 15 | |
| 16 | ## Current Plugins |
| 17 | |
| 18 | | Plugin | Type | Description | |
| 19 | |-----------------|-------|----------------------------------------------------------| |
| 20 | | `python-expert` | Agent | Python 3.11+, async/await, Pydantic, FastAPI, SQLAlchemy | |
| 21 | | `kotlin-expert` | Agent | Kotlin 2.0+, coroutines, Spring Boot, domain modeling | |
| 22 | | `react-expert` | Agent | React 19+, concurrent rendering, Tailwind, accessibility | |
| 23 | |
| 24 | ## Installation & Usage |
| 25 | |
| 26 | ### Add to Claude Code |
| 27 | |
| 28 | Add it to your project's `.mcp.json` (project-level) or `~/.claude/claude_mcp_settings.json` (global): |
| 29 | |
| 30 | ```json |
| 31 | { |
| 32 | "mcpServers": { |
| 33 | "provectus-claude-plugins-finder": { |
| 34 | "command": "npx", |
| 35 | "args": ["-y", "@provectusinc/claude-plugins@latest"] |
| 36 | } |
| 37 | } |
| 38 | } |
| 39 | ``` |
| 40 | |
| 41 | ### Using the plugins |
| 42 | |
| 43 | Once connected, three tools become available in your Claude Code sessions: |
| 44 | |
| 45 | | Tool | What it does | Example prompt | |
| 46 | |------------------|-------------------------------------------|-----------------------------------------------| |
| 47 | | `list_plugins` | Browse all plugins, filter by type or tag | *"List all available plugins"* | |
| 48 | | `get_plugin` | Retrieve a plugin's full content | *"Get the python-expert agent prompt"* | |
| 49 | | `search_plugins` | Search plugins by keyword | *"Search for plugins related to code review"* | |
| 50 | |
| 51 | You can ask Claude naturally and it will call the right tool: |
| 52 | |
| 53 | - *"What plugins are available for Python?"* |
| 54 | - *"Show me the react-expert agent"* |
| 55 | - *"Find plugins tagged with 'git'"* |
| 56 | |
| 57 | Claude will use the retrieved plugin content (agent prompts, skills, etc.) to enhance its responses with specialized expertise. |
| 58 | |
| 59 | ### Add to other MCP clients |
| 60 | |
| 61 | Any MCP-compatible client can connect to the server via stdio. The server package is `@provectusinc/claude-plugins` (binary name: `claude-plugins`): |
| 62 | |
| 63 | ```bash |
| 64 | npx -y @provectusinc/claude-plugins |
| 65 | ``` |
| 66 | |
| 67 | Configure your client to spawn this command and communicate over stdin/stdout using the [MCP protocol](https://modelcontextprotocol.io). |
| 68 | |
| 69 | --- |
| 70 | |
| 71 | ## Current Repo |
| 72 | |
| 73 | This repository is both a collection of plugins and an MCP server. The server exposes tools that allow any MCP-compatible client — including other Claude Code sessions — to discover, search, and retrieve plugins programmatically. |
| 74 | |
| 75 | ### Writing a Good Agent |
| 76 | |
| 77 | Agent prompts in `agents/*.md` define a specialized expert. A strong agent prompt: |
| 78 | |
| 79 | - **States the role clearly** up front — what the agent is an expert in. |
| 80 | - **Encodes real opinions** — prefer pattern X over Y, always use Z. Generic advice is not useful. |
| 81 | - **Includes concrete patterns** — show the actual code structure you want, not just descriptions of it. |
| 82 | - **Covers failure modes** — what to watch out for, common mistakes, performance traps. |
| 83 | - **Stays focused** — one domain of expertise per agent. Compose multiple agents rather than building one that tries to cover everything. |
| 84 | |
| 85 | ### Writing a Good Skill |
| 86 | |
| 87 | Skill prompts in `skills/*/SKILL.md` define a slash command. A strong skill prompt: |
| 88 | |
| 89 | - **Defines the task** — what the skill does when invoked. |
| 90 | - **Specifies inputs** — what arguments or context it needs. |
| 91 | - **Describes the expected output** — format, structure, level of detail. |
| 92 | |
| 93 | ### Guidelines |
| 94 | |
| 95 | - Keep plugin names lowercase and hyphenated (`my-plugin`, not `MyPlugin`). |
| 96 | - Use descriptive keywords in the manifest — they power the search tool. |
| 97 | - One concern per plugin. If you're mixing unrelated expertise, split it into separate plugins. |
| 98 | - W |