Add MCPB packaging and Claude Code plugin for pdf-server (#491)

* Add MCPB packaging and Claude Code plugin for pdf-server

Two distribution paths for the pdf-server:

1. **MCPB bundle** (Claude Desktop MCP Directory):
   - manifest.json: v0.3 manifest with tools, compatibility, privacy
   - .mcpbignore: excludes source/tests from bundle
   - icon.png: 64x64 PDF document icon

2. **Claude Code plugin** (plugin/):
   - .mcp.json: declares local stdio server via npx
   - commands: /pdf-research:read, /pdf-research:summarize
   - skills: pdf-reading skill for document navigation
   - Tested end-to-end: MCP server connects, all tools available

The plugin uses a local stdio MCP server pattern — the first known
instance in a knowledge-work plugin (all existing plugins use remote
HTTP connectors).

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* Run prettier:fix to satisfy CI formatting checks

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

examples/pdf-server/.mcpbignore

@@ -0,0 +1,21 @@
+# Source files (built output in dist/ is what we need)
+*.ts
+tsconfig*.json
+vite.config.ts
+src/
+
+# Tests
+*.test.*
+
+# Development assets
+screenshot.png
+grid-cell.png
+pdf-icon.svg
+
+# Plugin directory (separate distribution path)
+plugin/
+
+# Git and CI
+.git/
+.github/
+.gitignore

examples/pdf-server/manifest.json

@@ -0,0 +1,48 @@
+{
+  "manifest_version": "0.3",
+  "name": "@modelcontextprotocol/server-pdf",
+  "display_name": "PDF (By Anthropic)",
+  "version": "1.0.1",
+  "description": "Read and interact with PDF files using an interactive viewer with search, navigation, and text extraction",
+  "author": {
+    "name": "Anthropic",
+    "url": "https://www.anthropic.com"
+  },
+  "homepage": "https://github.com/modelcontextprotocol/ext-apps/tree/main/examples/pdf-server",
+  "repository": {
+    "type": "git",
+    "url": "https://github.com/modelcontextprotocol/ext-apps"
+  },
+  "support": "https://github.com/modelcontextprotocol/ext-apps/issues",
+  "icon": "icon.png",
+  "server": {
+    "type": "node",
+    "entry_point": "dist/index.js",
+    "mcp_config": {
+      "command": "node",
+      "args": ["${__dirname}/dist/index.js", "--stdio"],
+      "env": {}
+    }
+  },
+  "tools": [
+    {
+      "name": "list_pdfs",
+      "description": "List available PDF files that can be displayed"
+    },
+    {
+      "name": "display_pdf",
+      "description": "Display an interactive PDF viewer with search and navigation"
+    },
+    {
+      "name": "read_pdf_bytes",
+      "description": "Read a range of bytes from a PDF file (max 512KB per request)"
+    }
+  ],
+  "keywords": ["pdf", "documents", "viewer", "reading"],
+  "license": "MIT",
+  "compatibility": {
+    "platforms": ["darwin", "win32", "linux"],
+    "runtimes": { "node": ">=18.0.0" }
+  },
+  "privacy_policies": ["https://www.anthropic.com/privacy"]
+}

examples/pdf-server/plugin/.claude-plugin/plugin.json

@@ -0,0 +1,8 @@
+{
+  "name": "pdf-research",
+  "version": "0.1.0",
+  "description": "Read, navigate, and analyze PDF documents — academic papers, reports, and local files. Uses a local PDF MCP server for interactive viewing and text extraction.",
+  "author": {
+    "name": "Anthropic"
+  }
+}

examples/pdf-server/plugin/.mcp.json

@@ -0,0 +1,8 @@
+{
+  "mcpServers": {
+    "pdf": {
+      "command": "npx",
+      "args": ["@modelcontextprotocol/server-pdf", "--stdio"]
+    }
+  }
+}

examples/pdf-server/plugin/CONNECTORS.md

@@ -0,0 +1,16 @@
+# Connectors
+
+## Local MCP Server
+
+This plugin uses a **local MCP server** instead of a remote connector.
+The PDF server runs on your machine via `npx`.
+
+| Category   | Server                             | How it runs                            |
+| ---------- | ---------------------------------- | -------------------------------------- |
+| PDF viewer | `@modelcontextprotocol/server-pdf` | Local stdio via `npx` (auto-installed) |
+
+### Requirements
+
+- Node.js >= 18
+- Internet access for remote PDFs (arXiv, bioRxiv, etc.)
+- No API keys or authentication needed

examples/pdf-server/plugin/README.md

@@ -0,0 +1,40 @@
+# PDF Research Plugin
+
+Read and analyze PDF documents using a local MCP server. Designed for
+academic paper review, document analysis, and research workflows.
+
+## What It Does
+
+- **Open PDFs** from local files, arXiv, bioRxiv, and other academic sources
+- **Interactive viewer** with search, navigation, and zoom
+- **AI-powered analysis** -- summarize, extract key points, compare papers
+
+## Commands
+
+| Command                   | What it does           |
+| ------------------------- | ---------------------- |
+| `/pdf-research:read`      | Open a PDF for reading |
+| `/pdf-research:summarize` | Summarize a document   |
+
+## How It Works
+
+This plugin uses a **local MCP server** (`@modelcontextprotocol/server-pdf`)
+that runs on your machine via `npx`. No API keys or remote services needed --
+the PDF server starts automatically when the plugin loads.
+
+This is different from other knowledge-work plugins which use remote HTTP
+connectors. The local server pattern enables offline PDF access and avoids
+the need for a remote deployment.
+
+## Requirements
+
+- Node.js >= 18
+- Internet for remote PDFs (arXiv, etc.)
+
+## Supported PDF Sources
+
+- Local files (file paths in your working directory)
+- [arXiv](https://arxiv.org) papers
+- [bioRxiv](https://biorxiv.org) / [medRxiv](https://medrxiv.org) preprints
+- [chemRxiv](https://chemrxiv.org), [Zenodo](https://zenodo.org), [OSF](https://osf.io)
+- [HAL Science](https://hal.science), [SSRN](https://ssrn.com)

examples/pdf-server/plugin/commands/read.md

@@ -0,0 +1,21 @@
+---
+description: Open a PDF for reading and analysis
+---
+
+# Read PDF
+
+Open a PDF document for interactive reading. Accepts local file paths,
+arXiv URLs, or other academic paper URLs.
+
+## Instructions
+
+1. If the user provides a URL or file path, call `display_pdf` with that URL
+2. If no URL given, call `list_pdfs` first to show available documents
+3. After displaying, offer to summarize, extract key points, or answer questions
+
+## Supported Sources
+
+- Local files (drag-and-drop or file path)
+- arXiv papers (arxiv.org/abs/... or arxiv.org/pdf/...)
+- bioRxiv, medRxiv, chemRxiv preprints
+- Zenodo, OSF, HAL, SSRN

examples/pdf-server/plugin/commands/summarize.md

@@ -0,0 +1,23 @@
+---
+description: Summarize the current PDF or a new document
+---
+
+# Summarize PDF
+
+Generate a structured summary of a PDF document.
+
+## Instructions
+
+1. If a PDF is already displayed, work with it
+2. If user provides a new URL/path, call `display_pdf` first
+3. Read through the document and produce:
+   - **Title and authors** (if academic paper)
+   - **Key findings** (3-5 bullet points)
+   - **Methodology** (brief summary)
+   - **Conclusions**
+   - **Relevance** — what makes this paper notable
+
+## Notes
+
+- For long documents, focus on abstract, introduction, and conclusion sections
+- Ask the user which sections they care most about if the doc is very long

examples/pdf-server/plugin/skills/pdf-reading/SKILL.md

@@ -0,0 +1,35 @@
+---
+name: pdf-reading
+description: Read and navigate PDF documents using the local PDF server. Reference this when the user wants to open, read, search, or interact with any PDF file or academic paper.
+---
+
+# PDF Reading
+
+You have access to a local PDF server that provides interactive document viewing.
+
+## Available Tools
+
+- **list_pdfs** -- Show available PDFs. Call with no arguments.
+- **display_pdf** -- Open a PDF in the interactive viewer.
+  - `url` (string): URL or local file path
+  - `page` (number, optional): Starting page number
+
+## How to Use
+
+**When the user mentions a PDF, paper, or document:**
+1. If they give a URL or path, call `display_pdf` directly
+2. If they say "open the paper" without specifying, call `list_pdfs` to show options
+3. After displaying, offer to summarize, extract data, answer questions
+
+**arXiv shortcuts:**
+- `arxiv.org/abs/2301.12345` is auto-converted to the PDF URL
+- Users can just say "open arxiv 2301.12345"
+
+**Supported remote sources:**
+arXiv, bioRxiv, medRxiv, chemRxiv, Zenodo, OSF, HAL Science, SSRN
+
+## Best Practices
+
+- Always display the PDF before trying to analyze it
+- For multi-page documents, ask which section the user cares about
+- When comparing papers, display them one at a time and note key differences