Boost Java Microservice Analysis with Easy Code Reader MCP

Introduction

Easy Code Reader is an MCP (Model Context Protocol) server that helps developers retrieve relevant Java source code from multiple projects or JAR packages based on call‑chain analysis, allowing AI code agents such as Claude or ChatGPT to understand logic without manually copying files.

Best‑Practice Scenarios

1. Cross‑project RPC analysis

In complex micro‑service architectures, a single business operation may span several services. By cloning the relevant code locally and invoking the Easy Code Reader MCP tools, developers can let the AI agent trace the implementation of an RPC method such as submitOrderPaaS. The article provides a concrete prompt that asks the agent to locate the method in YbServiceSubmitOrderRpcImpl.java and explain its logic.

@Slf4j
@Service
public class YbServiceSubmitOrderRpcImpl implements YbServiceSubmitOrderRpc {
    @Resource
    private YbServiceSubmitOrderPaasResource ybServiceSubmitOrderPaasResource;
    @Override
    public String newSubmitOrder(YbSubmitOrderReq submitOrderReq) {
        if (log.isInfoEnabled()) {
            log.info("[提交订单]请求开始，参数-submitOrderReq：" + JSON.toJSONString(submitOrderReq));
        }
        // ... method body omitted for brevity ...
    }
}

2. Reading JAR source code

When a dependency is packaged as a JAR, the AI cannot read its code directly. Easy Code Reader provides the read_jar_source tool, which first tries to extract the source from a -sources.jar and falls back to decompilation if necessary. An example prompt asks the agent to fetch the source of com.jd.xxx.OfflineInsureOrderResource without specifying the Maven coordinates; the tool infers the correct artifact from the local pom.xml.

3. Cross‑project source reading for implementation

For features that rely on code in another module, a prompt can request the AI to read the target project's files (using list_all_project, list_project_files, and read_project_code) and then generate the missing implementation in the current project.

Additional Use Cases

Quick exception root‑cause tracing by reading the offending JAR source.

Assessing impact of dependency upgrades by comparing old and new source versions.

Code‑review assistance across multiple repositories.

Accelerating onboarding for new developers by visualising call chains.

Environment Requirements

Python 3.10+ (install via uv or system package manager).

JDK 8+ for the built‑in decompiler.

Quick Start

Method 1 – Using uvx (recommended)

Configure the MCP client with the Maven repository path ( --maven-repo) and the root directory of all local projects ( --project-dir). Example JSON configuration:

{
  "mcpServers": {
    "Easy Code Reader": {
      "command": "uvx",
      "args": [
        "easy-code-reader",
        "--maven-repo",
        "/customer/maven-repository",
        "--project-dir",
        "/customer/path/to/project-dir"
      ]
    }
  }
}

Method 2 – Installing locally

If uvx fails, install the binary with uv tool install easy-code-reader, locate it (e.g., /Users/you/.local/bin/easy-code-reader), and reference the absolute path in the MCP configuration.

Tool Details

list_all_project

Lists every sub‑directory under the configured project root, optionally filtered by project_name_pattern. It helps the AI discover available services and validates project existence.

list_project_files

Returns all source and configuration files for a given project, supporting both full‑project and focused sub‑path modes. File‑name fuzzy matching ( file_name_pattern) speeds up locating specific classes.

read_project_code

Retrieves the full content of a specified file or Java class, automatically searching common source locations (e.g., src/main/java, src/main/resources). Supports multi‑module Maven/Gradle layouts.

read_jar_source

Extracts a class from a Maven artifact. By default it prefers the -sources.jar; if unavailable it decompiles the binary using CFR for Java 8‑20 or Fernflower for Java 21+. The response includes a source_type field indicating whether the code came from a sources JAR, a fresh decompilation, or a cached decompilation.

Implementation Overview

The four tools cooperate to enable the AI to:

Discover all local projects ( list_all_project).

Identify relevant files ( list_project_files).

Read the exact source content ( read_project_code or read_jar_source).

Smart hint mechanisms guide the AI when patterns are too strict, suggesting a broader query.

Images