Pinboard MCP Server

Read-only access to Pinboard.in bookmarks for LLMs via Model Context Protocol (MCP).

Overview

This server provides LLMs with the ability to search, filter, and retrieve bookmark metadata from Pinboard.in at inference time. Built on FastMCP, it offers five tools for bookmark interaction while respecting Pinboard's rate limits and implementing intelligent caching.

Features

• Read-only access to Pinboard bookmarks
• Five MCP tools: search_bookmarks, search_bookmarks_extended, list_recent_bookmarks, list_bookmarks_by_tags, list_tags
• Smart caching with LRU cache and automatic invalidation using posts/update endpoint
• Rate limiting respects Pinboard's 3-second guideline between API calls
• Field mapping converts Pinboard's legacy field names to intuitive ones (description→title, extended→notes)
• Result cards for MCP Apps hosts: bookmarks and tags render as interactive cards in Claude, ChatGPT and other hosts that support MCP Apps; other hosts see plain results as before
• Comprehensive testing with integration test harnesses and CI validation

Installation

Via pip (recommended)

pip install pinboard-bookmarks-mcp-server

From source

git clone https://github.com/rossshannon/pinboard-bookmarks-mcp-server.git
cd pinboard-bookmarks-mcp-server
pip install -e .

Quick Start

1. Get your Pinboard API token from https://pinboard.in/settings/password
2. Set environment variable:

   export PINBOARD_TOKEN="username:1234567890ABCDEF"
   

3. Start the server:

   pinboard-mcp-server
   

4. Verify it's working:

   # Test help command (works without token)
   pinboard-mcp-server --help
   
   # Server should show "Starting MCP server" when run with token
   

Usage with Claude Desktop

Add this configuration to your Claude Desktop settings:

{
  "mcpServers": {
    "pinboard": {
      "command": "pinboard-mcp-server",
      "env": {
        "PINBOARD_TOKEN": "your-username:your-token-here"
      }
    }
  }
}

Available Tools

1. search_bookmarks

Search bookmarks by query string across titles, notes, and tags. Recent-focused with automatic expansion.

Parameters:

• query (string): Search query
• limit (int, optional): Maximum results (default: 20, max: 100)

Example:

Search for "python testing" bookmarks

2. search_bookmarks_extended

Extended search for comprehensive historical results across titles, notes, and tags.

Parameters:

• query (string): Search query
• days_back (int, optional): How many days back to search (default: 365, max: 730)
• limit (int, optional): Maximum results (default: 100, max: 200)

Example:

Search the last 2 years for "kubernetes" bookmarks

3. list_recent_bookmarks

List bookmarks saved in the last N days.

Parameters:

• days (int, optional): Days to look back (default: 7, max: 30)
• limit (int, optional): Maximum results (default: 20, max: 100)

Example:

Show me bookmarks from the last 3 days

4. list_bookmarks_by_tags

List ALL bookmarks filtered by tags with optional date range. Most efficient for historical access.

Parameters:

• tags (array): List of tags to filter by (1-3 tags)
• from_date (string, optional): Start date in ISO format (YYYY-MM-DD)
• to_date (string, optional): End date in ISO format (YYYY-MM-DD)
• limit (int, optional): Maximum results (default: 100, max: 200)

Example:

Find bookmarks tagged with "python" and "api" from January 2024

5. list_tags

List all tags with their usage counts.

Example:

What are my most used tags?

Configuration

Environment Variables

• PINBOARD_TOKEN (required): Your Pinboard API token in format username:token
• PINBOARD_MCP_APPS_UI (optional): set to false to turn off the result cards; on by default

Result cards (MCP Apps)

Each tool links to a ui://pinboard/… card: a list of bookmarks (title, site, date, notes and tags, with the tags you searched for highlighted) for the four bookmark tools, and a bar chart of your most-used tags with a filter for list_tags. The cards follow the host's light or dark theme, use the host's fonts, open links through the host, and load nothing else from the network. Hosts without MCP Apps support ignore them.

The cards are React, built into self-contained HTML under src/pinboard_mcp_server/ui/generated/ and shipped with the package, so running the server needs no Node. See docs/mcp-apps-ui.md for how they work and how to change them.

Rate Limiting

The server automatically enforces a 3-second delay between Pinboard API calls to respect their guidelines. Cached responses are returned immediately.

Caching Strategy

• Query cache: LRU cache with 1000 entries for search results
• Bookmark cache: Full bookmark list cached for 1 hour
• Cache invalidation: Uses posts/update endpoint to detect changes
• Tag cache: Tag list cached until manually refreshed

Testing

The project includes comprehensive test coverage with multiple test strategies:

Run all tests

# Activate virtual environment first
source ~/.venvs/pinboard-bookmarks-mcp-server/bin/activate

# Run all tests with coverage
pytest --cov=src --cov-report=term-missing

Real API testing

# Set your Pinboard token
export PINBOARD_TOKEN="username:token"

# Run debug utility to test search functionality (development only)
PINBOARD_TOKEN="username:token" python tests/debug_bookmarks.py

Mock API testing

# Run comprehensive test suite (development only)
python -m pytest tests/ -v

Development

Setup

# Clone and setup
git clone https://github.com/rossshannon/pinboard-bookmarks-mcp-server.git
cd pinboard-bookmarks-mcp-server

# Quick development setup
./scripts/dev-setup.sh

Code Quality

# Activate environment
source ~/.venvs/pinboard-bookmarks-mcp-server/bin/activate

# Linting and formatting
ruff check src/ tests/
ruff format src/ tests/

# Type checking
mypy src/

# Run tests
pytest -v

# Build package
./scripts/build.sh

Architecture

• FastMCP: MCP scaffolding with Tool abstraction and async FastAPI server
• pinboard.py: Pinboard API client wrapper with error handling
• Pydantic: Data validation and serialization with JSON Schema
• ThreadPoolExecutor: Bridges async MCP with sync pinboard.py library
• LRU Cache: In-memory caching with intelligent invalidation

Key Files

• src/pinboard_mcp_server/main.py - MCP server entry point and tool implementations
• src/pinboard_mcp_server/client.py - Pinboard API client with caching
• src/pinboard_mcp_server/models.py - Pydantic data models
• src/pinboard_mcp_server/ui/ - MCP Apps card resources and tool metadata
• cards/ - Source, fixtures and tests for the result cards (Node; see docs/mcp-apps-ui.md)
• tests/ - Comprehensive test suite
• tests/debug_bookmarks.py - Debug utility for testing search functionality
• docs/TEST_HARNESS.md - Documentation for test harnesses

Performance

• P50 response time: <250ms (cached responses)
• P95 response time: <600ms (cold cache)
• Rate limiting: 3-second intervals between API calls
• Cache hit ratio: >90% for typical usage patterns

Security

• API tokens are never logged or exposed in error messages
• Read-only access to Pinboard data
• Input validation on all tool parameters
• Secure environment variable handling

Troubleshooting

Common Issues

"PINBOARD_TOKEN environment variable is required"

• Make sure you've set your token: export PINBOARD_TOKEN="username:token"
• Get your token from https://pinboard.in/settings/password
• Token format should be: username:1234567890ABCDEF

"Command not found: pinboard-mcp-server"

• Ensure you've installed the package: pip install pinboard-bookmarks-mcp-server
• Check your Python environment is activated
• Try reinstalling: pip uninstall pinboard-bookmarks-mcp-server && pip install pinboard-bookmarks-mcp-server

Server starts but Claude Desktop can't connect

• Verify the MCP configuration in Claude Desktop settings
• Check that the command path is correct: pinboard-mcp-server
• Ensure the PINBOARD_TOKEN is set in the env section

"Permission denied" or "Access denied" errors

• Verify your Pinboard token is valid and active
• Check you have internet connectivity to reach pinboard.in
• Test your token manually at https://pinboard.in/api/v1/posts/recent

Contributing

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Make your changes with tests
4. Ensure all tests pass and code is formatted
5. Submit a pull request

License

MIT License - see LICENSE file for details.