MVP: Implement core MCP server with documentation and source tools #1

New issue

Open

opened 2025-12-28 13:05:57 +00:00 by mprihoda · 0 comments

mprihoda commented

2025-12-28 13:05:57 +00:00

(Migrated from github.com)

Overview

Implement the initial MVP of javadocs-mcp - an HTTP-based MCP server that provides on-demand Javadoc/Scaladoc and source code for JVM dependencies.

Goals

HTTP-based MCP server (shared across projects, efficient caching)
Two tools: get_documentation and get_source
Class-level granularity
Support both Javadoc and Scaladoc
Use Coursier for artifact resolution

Scope

In Scope

HTTP MCP transport with SSE
get_documentation(coordinates, className) - fetch and return Javadoc/Scaladoc HTML
get_source(coordinates, className) - fetch and return source file content
Coursier integration for fetching -javadoc.jar and -sources.jar
Basic error handling (artifact not found, class not found)

Out of Scope (Future)

Markdown conversion of HTML docs
Project-aware dependency discovery (parsing build.sbt, pom.xml, etc.)
Method-level granularity
Package listing/navigation
Persistent caching (in-memory only for MVP)

Technical Decisions

Scala 3 with scala-cli
Coursier for dependency resolution
HTTP server choice TBD (Tapir/http4s vs simpler options)
MCP transport HTTP with SSE for streaming

Tasks

Detailed analysis and task breakdown needed for:

MCP HTTP transport implementation
Coursier integration
JAR content extraction
Tool interface design
Error handling strategy
Testing approach

Success Criteria

Claude Code can connect to the running server
Can successfully fetch Javadoc for a class like cats.effect.IO
Can successfully fetch source for the same class
Proper error messages for missing artifacts or classes

## Overview Implement the initial MVP of javadocs-mcp - an HTTP-based MCP server that provides on-demand Javadoc/Scaladoc and source code for JVM dependencies. ## Goals - HTTP-based MCP server (shared across projects, efficient caching) - Two tools: `get_documentation` and `get_source` - Class-level granularity - Support both Javadoc and Scaladoc - Use Coursier for artifact resolution ## Scope ### In Scope - HTTP MCP transport with SSE - `get_documentation(coordinates, className)` - fetch and return Javadoc/Scaladoc HTML - `get_source(coordinates, className)` - fetch and return source file content - Coursier integration for fetching `-javadoc.jar` and `-sources.jar` - Basic error handling (artifact not found, class not found) ### Out of Scope (Future) - Markdown conversion of HTML docs - Project-aware dependency discovery (parsing build.sbt, pom.xml, etc.) - Method-level granularity - Package listing/navigation - Persistent caching (in-memory only for MVP) ## Technical Decisions - **Scala 3 with scala-cli** - **Coursier** for dependency resolution - **HTTP server** choice TBD (Tapir/http4s vs simpler options) - **MCP transport** HTTP with SSE for streaming ## Tasks Detailed analysis and task breakdown needed for: 1. MCP HTTP transport implementation 2. Coursier integration 3. JAR content extraction 4. Tool interface design 5. Error handling strategy 6. Testing approach ## Success Criteria - Claude Code can connect to the running server - Can successfully fetch Javadoc for a class like `cats.effect.IO` - Can successfully fetch source for the same class - Proper error messages for missing artifacts or classes