How a Leading HR Tech Platform Cut Agent Complexity by 90% with AWS Agent Core

Protagona partnered with a leading HR technology company to migrate a legacy Bedrock Agent to AWS Agent Core Runtime, eliminating infrastructure overhead and establishing a foundation for multi-model cost optimization.

Industry

Startups & Software

Teams & Services

Tech & Tools

AWS Agent Core (Runtime + Gateway), Strands SDK, AWS Lambda (Node.js), DynamoDB, TypeSense, Cognito (JWT), CloudFormation infrastructure-as-code, Amplify CI/CD, Python (Agent Core Runtime container)

Key Data Points

Agent Core Runtime deployed with Strands agent operational

All 10 Lambda action groups converted to MCP-compatible format

Complete documentation and deployment handoff to client team

The Vision

A leading salary intelligence platform set out to deliver instant, accurate compensation insights across thousands of job titles at scale. The vision was clear: eliminate the infrastructure overhead and technical complexity holding their AI agent back, and replace it with a modern, serverless foundation built to grow with their product.

The Goal

The primary objective was to migrate an existing AWS Bedrock Agent to AWS Agent Core Runtime, achieving full feature parity without touching a single line of frontend code. The engagement needed to be completed in seven working days, dramatically reduce agent code complexity, unlock access to the latest Claude models, and establish a scalable, cost-efficient architecture ready for product launch.

The Challenge

An existing AWS Bedrock Agent was struggling to keep pace with a production-ready platform. Response times of 20 to 30 seconds, a single-model approach with no intelligent query routing, and unsustainable cost economics made it incompatible with a public product launch.

‍

Compounding the pressure, the engagement had to be completed in seven working days over Thanksgiving week, with zero frontend changes permitted, all existing infrastructure preserved, and a hard AWS funding deadline of December 15, 2025.

‍

The Solution

Protagona migrated the legacy Bedrock Agent to AWS Agent Core Runtime, deploying a Strands-powered serverless container that reduced agent code from 300 lines to just 30 while unlocking the latest Claude Sonnet 4 models and eliminating Fargate overhead entirely.

‍

Rather than rewriting existing Lambda functions, the team used Agent Core Gateway as an MCP proxy, preserving all 10 Lambda action groups without touching underlying code. Combined with full CloudFormation automation and Cognito JWT authentication, the entire solution was delivered, tested, and handed off within the seven-day timeline.

‍

Code Simplified at Scale

By migrating to AWS Agent Core Runtime with the Strands SDK, agent orchestration code was reduced from 300 lines to just 30, cutting complexity by 90% and significantly reducing the long-term maintenance burden for the client team.

‍

Serverless by Design

The move to a fully serverless architecture eliminated planned Fargate infrastructure costs entirely, replacing a fixed-cost model with a consumption-based approach that scales with demand and removes the overhead of ongoing infrastructure management.

‍

A Foundation Built to Grow

The migration established a multi-model routing framework with access to the full AWS Bedrock catalog, positioning the platform for 50 to 70% cost reductions through intelligent model selection and sub-10 second response times via future multi-agent architecture.