Rocket Close is a Detroit-based title agency and appraisal management company within Rocket Companies that provides title insurance, property valuation, and settlement services. As demand for mortgages and loans grew, title operations became a bottleneck in the homebuying process. Time-intensive, state-specific title examinations, combined with manual research and fragmented systems, slowed throughput and made it difficult for teams to keep pace with an expanding client base.

Title examiners must verify data from disparate sources. This requires searching through multiple systems, state guides, and county requirements. Local rules around probate or tax IDs further complicate their work. For example, a title examiner seeking to understand a county-specific recording requirement might spend hours navigating multiple sources.

To address these challenges, Rocket Close created Supercharger in collaboration with AWS. Supercharger is an agentic AI solution designed to reduce friction in the lending and homebuying process and optimize title operations workflows. It combines title and closing knowledge to guide teams through the order processing workflow, dynamically interacting with internal operations teams in natural language. By centralizing knowledge and automating research-heavy tasks, the solution generates actionable insights about orders, improves efficiency, and reduces the time spent searching for information. Ultimately, it enhances both operational efficiency and client experience.

In this post, we explore how Rocket Close built a solution using Strands Agents, large language models (LLMs), Amazon Bedrock, Amazon Bedrock Knowledge Bases, and Model Context Protocol (MCP) tools. We cover solution features, the rationale for the technology stack, lessons learned, and the business impact at Rocket Close.

The Supercharger solution is powered by Strands Agents, an open source agent harness SDK by AWS for building agents using the Anthropic Claude Large Language Model (LLM) through Amazon Bedrock, giving it the flexibility to choose different LLMs as the title assistants evolve. From a security perspective, the solution combines Amazon Bedrock Guardrails with row-level data entitlements to help prevent accidental access to customer-sensitive data through intelligent access controls. Conversations are logged with complete audit trails to meet compliance requirements. It integrates with Rocket Close operational databases containing order information, standard procedures, and policies for state-level title exams. The following diagram shows the six interconnected capabilities of Supercharger.

At the core of the Supercharger solution is a domain-specific agent driving conversation with Operations teams through six interconnected capabilities that work together to streamline the homeownership process. Conversation Analytics enables natural language processing that understands context and intent across multi-turn conversations, making interactions feel intuitive and human-like rather than rigid and transactional. Building on this conversational intelligence, state-level title examination assistance provides comprehensive checklists and guidance tailored to specific title examination requirements, providing teams with the right information at the right moment. The solution’s API-based integration connects with existing systems to maintain data consistency and avoid manual data entry, reducing errors and freeing teams to focus on high value work. Guardrails and Response Accuracy verify that every response meets quality standards and complies with regulatory requirements, protecting both the company and its clients. Comprehensive logging and monitoring provide complete visibility into system performance and user interactions, with full audit trails that meet compliance requirements. Finally, unified access to multiple data sources maintains complete context for decision-making, pulling together information that previously required checking multiple systems, creating unified experience for operations teams navigating complex title workflows.