Aldo Lushkja / blog

Introducing zdtp-mcp: A Lightweight, Zero-Framework Model Context Protocol Server for Targetprocess

4 min read
#mcp#java#targetprocess#ai#opensource#docker#graalvm

Managing project management tools like Apptio Targetprocess often involves navigating dense web interfaces, updating task statuses, and searching for entity relations across sprints.

With the rise of AI coding agents and assistants—such as Claude Code CLI, Gemini CLI, Cursor, and Claude Desktop—software engineers need a seamless bridge to inspect, create, and update Targetprocess backlog items directly from their terminal or IDE.

To solve this, I built zdtp-mcp: a lightweight, zero-framework Model Context Protocol (MCP) server for Targetprocess written in pure Java 25.

Here is a dive into why zdtp-mcp was created, its architecture, and how you can get started in seconds.

What is Model Context Protocol (MCP)?

The Model Context Protocol (MCP) is an open standard introduced by Anthropic that enables AI models to safely interface with external tools, APIs, and databases. Instead of writing custom API integration scripts for every AI client, an MCP server exposes a standardized set of tools, resources, and prompts over stdio or HTTP/SSE.

By adding zdtp-mcp to your AI client, your AI assistant can execute natural language prompts like:

  • “Find all high-priority open bugs assigned to me in Sprint 42”
  • “Create a User Story for the payment gateway integration under Epic EP-102”
  • “Link User Story US-456 as a blocker for US-789”
  • “Generate a test plan with test steps for the auth service refactor”

Why Pure Java 25 & Zero Frameworks?

Many enterprise MCP servers are built using bulky frameworks like Spring Boot or Quarkus, or reliance on heavy Node.js runtimes. For a CLI tool or background daemon, framework overhead introduces delayed cold-start times and high memory usage.

zdtp-mcp was designed from the ground up with a zero-framework philosophy:

  1. Near-Instant Cold Start: Zero reflection scanning, no heavy dependency injection containers. Starts in milliseconds.
  2. Pure Java 25: Utilizes modern Java 25 features (virtual threads, pattern matching, records) for clean, high-concurrency standard I/O and HTTP processing.
  3. GraalVM Native Image Support: Can be compiled into a standalone, single-file native executable (zdtp-mcp) requiring zero JVM installation on host machines.
  4. Low Memory Footprint: Operates comfortably under 30MB of RAM in Docker containers.

Architecture: How zdtp-mcp Connects AI Agents to Targetprocess

![zdtp-mcp Architecture Sequence Diagram](https://mermaid.ink/svg/c2VxdWVuY2VEaWFncmFtCiAgICBhdXRvbnVtYmVyCiAgICBhY3RvciBVc2VyIGFzIFNvZnR3YXJlIEVuZ2luZWVyCiAgICBwYXJ0aWNpcGFudCBBSSBhcyBBSSBBc3Npc3RhbnQgKENsYXVkZSAvIEdlbWluaSkKICAgIHBhcnRpY2lwYW50IE1DUCBhcyB6ZHRwLW1jcCAoSmF2YSAyNSkKICAgIHBhcnRpY2lwYW50IFRQIGFzIFRhcmdldHByb2Nlc3MgUkVTVCBBUEkKCi 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)

View Mermaid Source Code 
sequenceDiagram
    autonumber
    actor User as Software Engineer
    participant AI as AI Assistant (Claude / Gemini)
    participant MCP as zdtp-mcp (Java 25)
    participant TP as Targetprocess REST API

    User->>AI: "Find all open bugs in Sprint 42"
    AI->>MCP: 1. JSON-RPC method "callTool" (bugs_search)
    activate MCP
    MCP->>TP: 2. HTTP GET [/api/v1/Bugs?where=...] (API Token Auth)
    TP-->>MCP: 3. Return Targetprocess JSON Response
    MCP-->>AI: 4. Formatted MCP Tool Result (Text / JSON)
    deactivate MCP
    AI-->>User: 5. Present Summarized Bug List with Links

Comprehensive Toolset (52 Tools Across 15 Domains)

zdtp-mcp exposes 52 specialized tools covering complete CRUD and relational operations:

Domain Available Operations
User Stories search, create, update, get, delete
Tasks search, create, update, get, delete
Bugs search, create, update, get, delete
Epics & Features search, create, update, get, delete
Releases & Sprints search, get, create, update, delete
Test Plans & Cases search, create, update, get, delete, add_step, delete_step
Relations & Links search, create (Blocker, Relation, Dependency), delete
Comments & History add_comment, list_comments

🚀 Quick Start Guide

1. Claude Code CLI

Run zdtp-mcp instantly using Docker:

claude mcp add zdtp -- docker run -i --rm \
  -e TP_URL="https://youraccount.tpondemand.com" \
  -e TP_TOKEN="your_api_token" \
  ghcr.io/aldo-lushkja/zdtp-mcp:latest

2. Gemini CLI

gemini mcp add zdtp docker run -i --rm \
  -e TP_URL="https://youraccount.tpondemand.com" \
  -e TP_TOKEN="your_api_token" \
  ghcr.io/aldo-lushkja/zdtp-mcp:latest

3. Claude Desktop Configuration (claude_desktop_config.json)

{
  "mcpServers": {
    "zdtp": {
      "command": "docker",
      "args": [
        "run", "-i", "--rm",
        "-e", "TP_URL=https://youraccount.tpondemand.com",
        "-e", "TP_TOKEN=your_api_token",
        "ghcr.io/aldo-lushkja/zdtp-mcp:latest"
      ]
    }
  }
}

Building Native Binaries with GraalVM

For environments where Docker overhead is unwanted, zdtp-mcp supports direct GraalVM compilation into a native binary:

# Clone the repository
git clone https://github.com/aldo-lushkja/zdtp-mcp.git
cd zdtp-mcp

# Build native binary with Gradle & GraalVM Native Image
./gradlew nativeBuild

# Run the native binary directly
./build/native/zdtp-mcp

Open Source & Contributing

zdtp-mcp is completely open source under the MIT license. Check out the repository on GitHub to contribute, report issues, or star the project:

👉 GitHub Repository: aldo-lushkja/zdtp-mcp