OpenClaw Java MVP

An experimental Java conversion (MVP) of the OpenClaw Personal AI Assistant.

Overview

This project implements a minimal viable version of OpenClaw in Java, focusing on:

• WebSocket Gateway (port 18789)
• Anthropic-powered Agent Runtime
• Console-based Interaction Channel
• Session Management with JSONL persistence

Prerequisites

• Java 21+ (virtual threads required)
• Anthropic API Key (sk-ant-...)

Getting Started

1. Build the project:

   ./gradlew build

2. Configure: Create ~/.openclaw-java/config.json:

   {
     "gateway": { "port": 18789, "authToken": "test-token" },
     "agent": { "provider": "anthropic", "apiKey": "sk-ant-...", "model": "claude-sonnet-4-20250514" }
   }

3. Run Gateway:

   java -jar build/libs/openclaw-java.jar gateway

4. Interact: The gateway will launch an interactive console prompt. Type messages to chat with the agent.

Project Structure

• src/main/java/ai/openclaw/gateway - WebSocket server
• src/main/java/ai/openclaw/agent - Agent logic loop & LLM provider
• src/main/java/ai/openclaw/channel - Console channel implementation
• src/main/java/ai/openclaw/config - Configuration loader
• src/main/java/ai/openclaw/session - Session storage

Running with Docker

You can run the application in a Docker container for an isolated environment.

1. Build the Docker image:

   docker build -t openclaw-java .

2. Run the container: You must provide your Anthropic API key as an environment variable.

   docker run -it --rm \
     -e ANTHROPIC_API_KEY=sk-ant-... \
     -p 18789:18789 \
     openclaw-java

   The container runs as a non-root user (openclaw) for security.

   • Code execution is confined to /home/openclaw/workspace.
   • File access is restricted to the workspace directory.
   • Network access to internal/private IPs is blocked.