Shopping for a house is without doubt one of the greatest monetary choices most individuals face, and LendingTree constructed a multi-agent mortgage assistant on Amazon Bedrock to make the method extra simple. The assistant educates debtors, understands their state of affairs, and gives tailor-made choices in a pure dialog. Debtors should weigh buy or refinance, typical or government-backed, 15-year or 30-year phrases, and glued or adjustable charges. On high of that, there’s jargon like “low cost factors,” “origination charges,” and “debt-to-income ratio.” It’s no marvel many individuals really feel misplaced earlier than they even begin.
LendingTree has been serving to customers type by these selections for over 25 years, connecting tens of millions of individuals with lenders to search out aggressive mortgage gives. The corporate was constructed on a perception that everybody deserves the instruments and data to make assured monetary choices. The AI-powered mortgage assistant is that subsequent step.
The answer needed to meet the identical requirements which have guided LendingTree from day one: correct info, clear steering, and rigorous safety of person information. Working inside the regulatory necessities of the mortgage business means content material filtering, personally identifiable info (PII) safety, and compliance oversight aren’t elective options. They’re non-negotiable. That made Amazon Bedrock and its built-in guardrails the best basis.
“Our aim was to be a trusted companion within the home-buying journey, a information that educates customers, understands their state of affairs, and matches them with the best provide. The muse fashions and built-in guardrails in Amazon Bedrock allow us to ship that with safety and compliance from the beginning.”
— Srinivas Madabushi, SVP Expertise, LendingTree
Many corporations within the business have added chatbots for primary questions. LendingTree wished to go additional, answering the exhausting questions and matching debtors with aggressive gives. That took a couple of agent.
Answer overview
To handle these challenges, LendingTree deployed three unbiased AI brokers: a supervisor and two specialised staff (schooling and matching), coordinated utilizing LangGraph, the Model Context Protocol (MCP), and basis fashions powered by Amazon Bedrock. All brokers run containerized companies on Amazon Elastic Container Service (Amazon ECS) with AWS Fargate.
LendingTree selected Amazon Bedrock for its multi-model flexibility and inherited AWS governance controls, which their compliance workforce required. The answer was deployed on Amazon ECS as an alternative of Amazon Bedrock AgentCore as a result of it was already in manufacturing when AgentCore reached basic availability. AgentCore is a platform to construct, join, and optimize brokers at scale, with any framework or mannequin.
The next determine reveals the high-level structure.
LendingTree multi-agent structure for mortgage steering
(1) The buyer-facing chat interface is a React software working as a containerized service on Amazon ECS with Fargate, serving customers on net and cellular browsers by a public endpoint.
(2) Person inputs and mannequin outputs move by Amazon Bedrock Guardrails for content material filtering (resembling hate and profanity detection) and PII redaction. Incoming messages are additionally screened for immediate threats.
In parallel with Guardrails, a security classifier primarily based on a big language mannequin (LLM) enforces LendingTree’s conversational coverage. The 2 security checks run concurrently, so this added assurance doesn’t add latency.
A business-logic layer handles LendingTree’s operational guidelines: routing advanced points to human help and redirecting off-topic conversations.
(3) The Supervisor agent is the orchestrator, constructed on LangGraph as a state machine that follows a plan-and-execute sample. It really works like a well-organized workforce: the Supervisor understands the large image and decides what must occur, whereas the employees deal with what they’re every finest at.
The Supervisor is a graph of nodes and edges. Nodes do the work (intent evaluation, execution planning, response composition), and edges determine what occurs subsequent primarily based on the outcome (path to Training, path to Matching, or reply straight). This makes the circulation specific: each path by the system is outlined, auditable, and traceable.
The Supervisor analyzes person intent with Amazon Nova Professional and produces an execution plan. It then routes to the suitable employee over MCP utilizing connection pooling for low-latency calls. For mannequin availability by Area, see Supported fashions by AWS Area in Amazon Bedrock.
Not each process wants probably the most highly effective mannequin. The Supervisor implements multi-model structure, routinely deciding on between Amazon Nova Professional (for advanced reasoning and important classification) and Amazon Nova Lite (for conversational responses and light-weight classification) relying on what the duty calls for. This balances reliability with value.
Planning and execution are separated by design. When a dialog goes fallacious, the workforce can hint precisely which node made which determination.
(4) The Training employee is the affected person educator. It helps customers perceive mortgage ideas, explains product variations, and gives steering by interactive dialog. It runs its personal LangGraph workflow and maintains specialised Amazon Bedrock Data Bases, the totally managed retrieval augmented era (RAG) functionality, tailor-made to its area, backed by Amazon OpenSearch Service because the vector retailer. Each response is grounded in actual paperwork fairly than counting on mannequin data alone.
(5) The Matching employee is the connector. It gathers person preferences and calls LendingTree’s inside provide, eligibility, and charge APIs. From there, it delivers customized lending choices primarily based on the person’s precise state of affairs. It interprets qualification standards and helps customers evaluate choices facet by facet.
(6) Shared companies deal with dialog reminiscence and session state by the LangGraph PostgreSQL checkpointer on Amazon Relational Database Service (Amazon RDS). This lets conversations persist throughout turns, agent handoffs, and repair restarts. Customers can pause, come again later, or ask follow-ups with out shedding context.
(7) Amazon Bedrock gives a broad set of basis fashions suited to a variety of use instances. These vary from advanced reasoning and important classification (Amazon Nova Professional) to cost-efficient conversational responses and light-weight classification (Amazon Nova Lite), all accessible by a single API. Constructed-in Guardrails ship configurable security controls, whereas Amazon Bedrock Data Bases join brokers to authoritative content material by vector shops resembling Amazon OpenSearch Service for hybrid retrieval.
(8) Inner LendingTree companies and APIs present lender search, charge and product information, eligibility and prequalification checks, provide comparability, and person profile info.
(9) All companies run on Amazon ECS with AWS Fargate, the place every agent scales independently by itself demand alerts. Infrastructure is managed by Terraform and deployed by GitLab steady integration and steady supply (CI/CD) pipelines with automated testing and well being checks.
Following a dialog finish to finish
Suppose a person asks, “What’s the distinction between FHA and traditional, and which one suits me?”
The message flows by Amazon Bedrock Guardrails (2) for security screening. The Supervisor agent (3) then masses dialog historical past from dialog reminiscence (6) and analyzes intent, recognizing two wants: a proof and a personalised comparability. It routes the academic query to the Training employee (4) over MCP, and the matching query to the Matching employee (5). Every employee operates independently and returns its outcome. The Supervisor aggregates each into one coherent reply, working the output again by Guardrails earlier than delivering it. The complete change is checkpointed to PostgreSQL (6), so the following flip picks up with full context.
To the person, it’s one steady dialog. Behind the scenes, it’s a coordinated workflow the place each determination level is traceable.
Deployment and operations
As a result of the brokers talk by MCP, every could be up to date, scaled, and rolled again independently. The Training employee can ship a brand new data base with out touching the Supervisor or the Matching employee. Operationally, the problem distinctive to multi-agent methods is tracing a single dialog that spans a number of companies. Amazon CloudWatch logs and AWS X-Ray distributed tracing enable the workforce to observe a single dialog’s journey throughout all three brokers, with per-agent metrics and detailed timing.
Outcomes
The assistant has been dealing with actual mortgage conversations in manufacturing since late 2025. The metrics on this part are primarily based on LendingTree’s inside manufacturing analytics from launch by Q1 2026, as reported by the LendingTree engineering workforce. Throughout that interval, it served roughly 1,960 conversations and 12,100 messages, averaging 6.2 messages per change. These aren’t one-shot lookups: engaged customers maintain multi-turn periods averaging 10+ messages over 9 minutes, asking follow-up questions, refining their state of affairs, and exploring choices conversationally.
From schooling to motion. The dialog information reveals how customers truly use an AI mortgage assistant. Early within the rollout, 75% of conversations have been academic. Customers requested questions like “What’s an FHA mortgage?”, “How does a HELOC work?”, or “What credit score rating do I want for a standard mortgage?” Because the system matured and phrase unfold, transactional intent grew: over 50% of latest conversations now contain charge comparisons, lender matching, or prequalification, actions tied on to conversion.
Dialog depth alerts belief. The common session size of 6.2 messages tells one story. The engaged-user common of 10+ messages over 9 minutes tells a unique one. Customers don’t maintain that sort of engagement with a system they don’t discover helpful. The Training employee builds understanding, and customers naturally progress to the Matching employee once they’re able to act. No pressured handoff, no “begin over” second.
The questions reveal actual complexity. The commonest matters embody mortgage sort comparisons (FHA, typical, and VA), qualification standards for particular credit score profiles, charge negotiation methods, closing timelines, and down cost trade-offs. These are precisely the situation-dependent questions that static FAQ pages can’t reply effectively. They require context: what the person mentioned three turns in the past, what they qualify for, what they’re optimizing for. That is the place dialog reminiscence and the Supervisor’s intent evaluation earn their complexity, and the system maintains the thread so every reply builds on the final.
Containment. Over 97% of conversations have been dealt with end-to-end with out human escalation, and solely about 3% of customers explicitly requested a stay agent. For a regulated monetary product the place the questions are genuinely exhausting and getting it fallacious has actual penalties, that containment charge is important. It demonstrates the system’s skill to function as a self-contained advisory service fairly than a triage layer in entrance of a name heart.
What the information confirms. The engagement depth is the sign that issues most. Higher-informed customers make higher choices. A conversational interface surfaces preferences and constraints {that a} static kind by no means captures together. Particulars like “I’m a veteran with a 650 credit score rating trying to purchase in Colorado Springs within the subsequent 30 days” make downstream lender matches extra related. That’s finally what serves the patron.
What LendingTree discovered, and what’s subsequent
Getting a multi-agent system to manufacturing taught the workforce as a lot about structure as about any single function.
After transport the mortgage assistant, LendingTree realized that constructing one agent is easy, however scaling to many brokers requires shared foundations. Relatively than constructing each new function as a one-off, LendingTree is investing in reusable items. These embody shared context layers for information entry, MCP contracts for instrument integration, and standardized deployment processes. A registry lets groups uncover current capabilities as an alternative of reinventing them. This lets groups construct smaller, domain-focused capabilities that may be composed into broader client experiences. Constant requirements for security, observability, analysis, and reuse tie all of it collectively.
Agent design learnings:
- Semantic chunking for Data Bases. Breaking paperwork into semantically coherent chunks fairly than fixed-size chunks improved retrieval high quality considerably, as a result of chunk boundaries align with pure subject breaks.
- Data Base battle decision. With a number of KBs, contradictory info typically surfaced. Area-based filtering and supply prioritization solved this: inside LendingTree content material takes priority for product-specific questions, whereas exterior assets serve basic mortgage schooling.
- Inter-agent context passing. Employee brokers initially lacked consciousness of the broader dialog. Passing full dialog historical past and intent summaries in every MCP request gave staff the context they should present related responses.
- Question rewriting. Quick person responses like “undecided” or “sure” get rewritten into significant, searchable queries utilizing dialog historical past earlier than retrieval. This considerably improved retrieval high quality.
- Guardrail tuning is ongoing work. Early configurations blocked authentic questions as a result of mortgage terminology tripped content material filters. Tuning in opposition to reasonable dialog information resolved this.
- Job-based mannequin routing stored prices in test: Nova Professional solely the place reasoning demanded it, Nova Lite all over the place else.
These are agent design issues, not infrastructure issues. They continue to be vital no matter runtime alternative, and characterize the lasting data from this implementation.
Infrastructure learnings:
- Dialog state administration. Early variations misplaced context throughout agent handoffs. A unified PostgreSQL-backed checkpointer with specific state serialization solved this, permitting conversations to persist throughout agent transitions and system restarts.
- Working security checks in parallel preserved latency with out weakening safety.
- Impartial agent scaling required hand-wiring ECS containers, well being checks, and deployment pipelines per agent.
The workforce is now evaluating re-architecting onto AgentCore to dump undifferentiated infrastructure work. That may keep away from hand-wiring PostgreSQL checkpointers, ECS containers, and agent scaling, letting engineering effort shift from “hold the runtime alive” to “hold optimizing the agent logic and area data.”
Conclusion
LendingTree’s manufacturing multi-agent system demonstrates that AI brokers can serve customers in closely regulated industries, not solely as prototypes, however as manufacturing methods working 24/7. Amazon Bedrock (Nova Professional and Nova Lite, Data Bases, and Guardrails) offered the mannequin and security basis. LangGraph dealt with agent orchestration, and MCP dealt with agent-to-agent communication. Collectively, they ship customized mortgage steering at scale whereas sustaining strict compliance requirements.
For organizations constructing multi-agent methods: separate planning from execution so the system is debuggable. Make security a structural pillar fairly than a skippable step. Design from day one for the reusable capabilities that can allow you to scale the sample throughout your group.
To get began with multi-agent methods on AWS, discover the Amazon Bedrock documentation and Amazon Bedrock AgentCore for managed runtime capabilities.
In regards to the authors
