DCR MCP Server

A Model Context Protocol (MCP) server providing development and research utilities: git analysis, literature search, markdown/PDF conversion, and email drafting.

Table of Contents

• Quick Start
• Installation Options
• Configuration
• Tools Reference
  • 🔍 Git Summary
  • 🔬 Literature Search
  • 📝 Markdown Converter
  • 📄 PDF Generator
  • ✉️ Email Prompt
• Development

Quick Start

1. Install

   go install github.com/dictybase/dcr-mcp/cmd/server@latest
   

2. Configure - Add to your MCP configuration:

   {
       "mcpServers": {
           "dcr-mcp": {
               "command": "server",
               "env": {
                   "OPENAI_API_KEY": "your-api-key-here"
               }
           }
       }
   }
   

3. Use - The server provides these tools:

   • Git Summary - Analyze commit messages with AI
   • Literature - Fetch research papers by PMID/DOI
   • Markdown - Convert to HTML with GFM support
   • PDF - Generate PDFs from markdown
   • Email - Draft casual emails

Installation Options

Requirements: Go 1.23.8+, OpenAI API key (for git summaries)

Configuration

MCP Setup

Use the correct command based on your installation method:

{
    "mcpServers": {
        "dcr-mcp": {
            "command": "server",  // for go install
            // or "command": "dcr-mcp-server",  // for source build
            // or "command": "/full/path/to/binary",  // for custom path
            "env": {
                "OPENAI_API_KEY": "your-api-key-here"
            }
        }
    }
}

The server auto-starts with your MCP client and provides structured logging to stderr for debugging.

Tools Reference

🔍 Git Summary

Analyzes git commits and generates AI-powered summaries categorized by type.

Parameters:

• repo_url (required) - Git repository URL
• branch (required) - Branch to analyze
• start_date (required) - Start date for analysis
• end_date (optional) - End date (defaults to now)
• author (required) - Filter by author name
• api_key (required) - OpenAI API key

Output: Markdown summary with categorized bullet points (features, bugs, docs, etc.)

🔬 Literature Search

Fetches scientific papers using PubMed IDs or DOIs via dictyBase API.

Parameters:

• id (required) - PMID (e.g., "12345678") or DOI (e.g., "10.1038/nature12373")
• id_type (required) - "pmid" or "doi"
• provider (optional) - "pubmed" (default) or "europepmc"

Output: Formatted article metadata with title, authors, abstract, and citation data

📝 Markdown Converter

Converts Markdown to HTML with GitHub Flavored Markdown support.

Parameters:

• content (required) - Markdown content to convert

Features: GFM, syntax highlighting, emoji, tables, YAML metadata

📄 PDF Generator

Converts Markdown to PDF with professional formatting.

Parameters:

• content (required) - Markdown content
• filename (optional) - Output filename (defaults to "output.pdf")

Output: PDF file with IBM Plex fonts and confirmation message

✉️ Email Prompt

Generates casual email drafts with customizable tone.

Parameters:

• from (required) - Sender's name or email
• to (required) - Recipient's name or email

Output: Complete email draft with subject line and friendly, professional tone

Development

Build Commands

Project Structure

dcr-mcp/
├── cmd/server/              # Main entry point
├── pkg/
│   ├── tools/              # Core tool implementations
│   │   ├── gitsummary/     # Git analysis + OpenAI
│   │   ├── literaturetool/ # PubMed/DOI fetching  
│   │   ├── markdowntool/   # Markdown → HTML
│   │   └── pdftool/        # Markdown → PDF
│   ├── prompts/            # Email prompt logic
│   └── markdown/           # Shared markdown utilities
└── bin/                    # Built binaries

Contributing

1. Fork and create feature branch
2. Follow conventions in CLAUDE.md
3. Add tests for new functionality
4. Run: make fmt && make test && golangci-lint run
5. Submit pull request

Debugging

The server logs to stderr with prefixed messages:

• [git-summary], [literature], [markdown], [pdf-tool], [email-prompt]

Enable stderr output in your MCP client for debugging.