Building an AI Coding Assistant with Claude for Spring Boot, AWS EKS & PostgreSQL

A step-by-step guide to creating a domain-expert AI assistant powered by Claude (Anthropic), with optional OpenClaw agent framework.

Architecture Overview

Developer (You)
    ↓ (chat / IDE / Slack)
OpenClaw Agent Runtime  ←→  Claude API (Anthropic)
    ↓                            ↑
Vector DB (RAG)          Knowledge Base
(Pinecone/Weaviate)      (Spring Boot, EKS, PostgreSQL docs)

Step 1: Get Your Claude API Access

1. Sign up at Anthropic Console
2. Generate an API key
3. Choose a model (e.g., claude-3-opus or claude-sonnet)

Step 2: Build a Domain Knowledge Base

Curate documents so Claude has strong, specific knowledge:

Format: Split into small markdown chunks (~500–1000 tokens each) with metadata tags.

Step 3: Set Up RAG (Retrieval-Augmented Generation)

This is what gives Claude deep, accurate domain knowledge beyond its training.

Components

1. Choose a Vector Database: Pinecone, Weaviate, pgvector (you can use PostgreSQL itself!), or OpenSearch
2. Embed your documents using an embedding model (e.g., Cohere, OpenAI text-embedding-3, or Voyage AI)
3. At query time:
   • Embed the user’s question
   • Retrieve top-k relevant chunks from the vector DB
   • Inject them into Claude’s prompt as context

Sample RAG Flow (Python)

import anthropic

client = anthropic.Anthropic(api_key="YOUR_KEY")

def ask_assistant(question):
    # 1. Retrieve relevant docs from vector DB
    context_chunks = vector_db.search(embed(question), top_k=5)
    context = "\n---\n".join(context_chunks)

    # 2. Call Claude with context
    response = client.messages.create(
        model="claude-sonnet-4-20250514",
        max_tokens=2048,
        system="""You are an expert coding assistant specializing in:
- Spring Boot (including Security, Data JPA, Actuator)
- AWS EKS (deployment, scaling, networking, IAM)
- PostgreSQL DBA (performance tuning, indexing, replication, backup)
Use the provided context to give accurate, actionable answers.""",
        messages=[{
            "role": "user",
            "content": f"Context:\n{context}\n\nQuestion: {question}"
        }]
    )
    return response.content[0].text

Step 4 (Optional): Use OpenClaw for a Full Agent Experience

OpenClaw turns Claude into an autonomous, persistent AI agent that can:

• Execute shell commands, manage files, browse the web
• Maintain memory across sessions
• Work via Slack, Discord, WhatsApp, or terminal
• Run scheduled tasks (like a cron job)

Quick Setup

# Install OpenClaw
npm install -g openclaw
# or
curl -sSL https://openclaw.im/install.sh | bash

# Initialize with Claude
openclaw init --model claude --api-key YOUR_ANTHROPIC_KEY

# Add custom skills for your stack
openclaw skill add spring-boot-helper
openclaw skill add eks-deployer
openclaw skill add postgres-dba

You can write custom skills (a folder with code + SKILL.md) so your agent knows how to:

• Generate Spring Boot boilerplate
• Run kubectl commands against your EKS cluster
• Execute psql queries for DB administration

Step 5: Technology Stack Summary

Step 6: Deploy on AWS EKS

1. Containerize your assistant service (Dockerfile)
2. Deploy to your EKS cluster via Helm chart or Kubernetes manifests
3. Store the Anthropic API key in AWS Secrets Manager or K8s Secrets
4. Expose via ALB Ingress or internal service for your team

Step 7: Iterate & Improve

• Log unanswered questions → add missing docs to knowledge base
• Fine-tune the system prompt for your team’s coding standards
• Add more skills (CI/CD helpers, code review automation, DB migration scripts)

Quick Start Recommendation

References

• Anthropic API Docs
• RAG Patterns (Pinecone)
• Spring Boot Reference
• AWS EKS Best Practices
• PostgreSQL Documentation
• OpenClaw Official Site
• OpenClaw GitHub

Published: 2026-05-07