rubensgomes/javamcp

JavaMCP

A Python-based AI MCP (Model Context Protocol) server that provides AI coding assistant agents with rich contextual information about Java APIs found in Java source code files stored in configured Git repositories.

The information gathered from the Java source code files includes Javadocs, public method signatures, class hierarchies, and usage examples.

The intent is that, during the development of a Java application or library, we can incorporate other Java APIs from libraries defined in the “javamcp” MCP-configured Git repositories.

Features

• 4 MCP Tools for comprehensive Java API exploration
• 1 MCP Resource for accessing comprehensive project context
• Rich Context with Javadoc summaries and API documentation
• Git Repository Integration for automatic code discovery
• ANTLR4-based Parser for accurate Java 21+ source code analysis
• Comprehensive Indexing for fast API lookups
• Query Engine with flexible search and filtering

Installation

NOTE: instructions provided for Unix-based systems (Linux, macOS) only.

Prerequisites

• UNIX-based system (e.g., Linux, macOS)
• Python 3.13+

Configuration

• Create and edit ~/.config/javamcp/config.yml to add your Git repositories. For example:

# Configuration for JavaMCP Server

# Server configuration
server:
    mode: http # Server mode: "stdio" or "http"
    host: localhost  # Host for HTTP mode
    port: 8000  # Port for HTTP mode

# Repository configuration
repositories:
    # The default branch in the remote Git repository is cloned.  That is, 
    # the branch that remote's HEAD reference points to. Historically, this 
    # branch has most commonly been named master, but main has become an 
    # increasingly prevalent default in recent years, particularly on 
    # platforms like GitHub. Other names like trunk or development are also
    # used.

    urls:
        # TODO: Add your git repo URLs below.  For example:
        - https://github.com/apache/commons-lang.git
        - https://github.com/google/guava.git
    local_base_path: ./repositories  # Local directory for cloned repos

# Logging configuration
logging:
    level: DEBUG # Log level: DEBUG, INFO, WARNING, ERROR, CRITICAL
    format: "%(asctime)s - %(name)s - %(levelname)s - %(message)s"
    output: both # Output: "console", "file", or "both"
    file_path: /tmp/javamcp.log # Log file path (required if output is "file" or "both")

Usage

Running the Server

poetry run python -m javamcp --config config.yml

MCP Tools

JavaMCP provides 4 MCP tools for AI assistants:

1. Search Methods (search_methods)

Search for Java methods by name with full context.

Request:

{
    "method_name": "substring",
    "class_name": "String",
    "case_sensitive": false
}

Response:

• Method signature
• Javadoc summary and description
• Parameter descriptions
• Return type documentation
• Containing class information

2. Analyze Class (analyze_class)

Get complete information about a Java class.

Request:

{
    "fully_qualified_name": "java.lang.String",
    "repository_name": "jdk"
}

Response:

• Class-level Javadoc
• All methods with Javadocs
• All fields with Javadocs
• Inheritance hierarchy
• Annotations

3. Extract APIs (extract_apis)

Extract and index APIs from a Git repository.

Request:

{
    "repository_url": "https://github.com/apache/commons-lang.git",
    "branch": "main",
    "package_filter": "org.apache.commons.lang3",
    "class_filter": "String"
}

Response:

• List of extracted classes with Javadocs
• Total classes and methods indexed
• API summaries

4. Generate Usage Guide (generate_guide)

Generate usage guides for specific use cases.

Request:

{
    "use_case": "How to manipulate strings",
    "max_results": 10
}

Response:

• Relevant classes and methods
• Javadoc-based usage examples
• Formatted guide with context

MCP Resources

JavaMCP provides 1 MCP resource for accessing project-level information:

Project Context Resource (javamcp://project/{repository_name}/context)

Access comprehensive contextual information about a Java API project.

URI Pattern:

javamcp://project/{repository_name}/context

Parameters:

• repository_name: Repository name (extracted from Git URL, e.g., " commons-lang" for "https://github.com/apache/commons-lang.git")

Response includes:

• Repository Information: Name and URL
• Project Description: Generated overview with statistics
• README Content: Full README.md file content (if available)
• LLMs.txt Content: Context file for LLMs (if available)
• API Statistics: Total classes, methods, packages, and averages
• Package Summaries: Package-level information with class and method counts
• Top Classes: Most significant classes with Javadocs
• Documentation Coverage: Javadoc coverage metrics for classes and methods

Example Usage:

If you have indexed https://github.com/apache/commons-lang.git, access its context via:

javamcp://project/commons-lang/context

Response Structure:

{
    "repository_name": "commons-lang",
    "repository_url": "https://github.com/apache/commons-lang.git",
    "description": "# commons-lang\n\nRepository: https://github.com/apache/commons-lang.git...",
    "readme_content": "# Apache Commons Lang...",
    "llms_txt_content": null,
    "statistics": {
        "total_classes": 150,
        "total_methods": 2500,
        "total_packages": 15,
        "average_methods_per_class": 16.67
    },
    "packages": [
        {
            "name": "org.apache.commons.lang3",
            "class_count": 50,
            "method_count": 800,
            "classes": [
                "StringUtils",
                "ArrayUtils",
                "ObjectUtils"
            ]
        }
    ],
    "top_classes": [
        {
            "name": "StringUtils",
            "fully_qualified_name": "org.apache.commons.lang3.StringUtils",
            "package": "org.apache.commons.lang3",
            "summary": "Operations on String that are null safe",
            "method_count": 120,
            "type": "class"
        }
    ],
    "javadoc_coverage": {
        "class_documentation_rate": 95.5,
        "method_documentation_rate": 87.3,
        "documented_classes": 143,
        "total_classes": 150,
        "documented_methods": 2183,
        "total_methods": 2500
    }
}

Development

Running Tests

# Run all tests
poetry run pytest

# Run with coverage
poetry run pytest --cov=src/javamcp --cov-report=term-missing

# Run specific test module
poetry run pytest tests/models/

Code Quality

# Format code
poetry run black src/ tests/

# Lint
poetry run pylint src/javamcp

# Type checking
poetry run mypy src/

# Sort imports
poetry run isort src/ tests/

Architecture

javamcp/
├── models/          # Pydantic data models
├── config/          # Configuration management
├── repository/      # Git repository operations
├── parser/          # ANTLR4 Java parser
├── indexer/         # API indexing and query engine
├── context/         # Context building and formatting
├── tools/           # MCP tool implementations
├── server/          # MCP server
├── logging/         # Logging utilities
└── utils/           # Helper utilities

Test Coverage

• 247 tests across all modules
• 95%+ average coverage for production code
• Comprehensive unit and integration tests

Claude Code Commands

• Update Claude Code to the latest version:

  claude update
  

• List configure MCP servers:

  claude mcp list
  

• See Claude Code version:

  claude --version
  

• Start Claude Code in IDE mode:

  claude --verbose --debug ide
  

Development Setup

1. Clone the repository:

   git clone https://github.com/rubensgomes/javamcp
   cd javamcp
   

2. Follow the instructions in the cloned repository DEVSETUP.md file.

Author: Rubens Gomes