Loan MCP Server
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Loan MCP Server Get loan summary for reference LN-2025-00123"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Loan MCP Server
A Model Context Protocol (MCP) server that exposes read and analytics tools for a Loan Management System backed by Microsoft Dataverse. It can be consumed by MCP clients (Claude Desktop, Copilot Studio) over stdio or HTTP, and includes an optional WhatsApp assistant.
Features
12 loan tools: lookups, search, timeline, eligibility, officer workload and portfolio analytics.
Two transports: stdio (Claude Desktop) and Streamable HTTP (Copilot Studio / remote clients), JSON-RPC 2.0.
Optional WhatsApp assistant via Twilio + an LLM.
Microsoft Entra ID authentication (client credentials); Dataverse Web API v9.2.
TypeScript, Node.js 22+, structured logging, typed error handling.
Related MCP server: dataverse-mcp-server
Tools
Tool | Input | Description |
|
| Full loan summary |
|
| Status, eligibility, officer, review flag |
|
| Loans for a phone number |
|
| Applications for an applicant |
|
| Milestone timeline |
|
| Eligibility explanation |
|
| Required document checklist |
| – | Loans under manual review |
|
| An officer's assigned loans and workload |
|
| Loans by status |
|
| Loans assigned to an officer |
| – | Portfolio totals and amount statistics |
Prerequisites
Node.js 22 or later
A Microsoft Dataverse environment
An Entra ID app registration with a corresponding Dataverse application user (Read access to the loan table)
Getting started
npm install
cp .env.example .env # then fill in the required values
npm run build
npm start # HTTP server (dist/http.js)For local development with reload:
npm run dev:http # HTTP
npm run dev # stdioConfiguration
Set these in .env (local) or as environment variables / Azure App settings (hosted).
Required
Variable | Description |
| Entra ID directory (tenant) ID |
| App registration (client) ID |
| Client secret value |
| e.g. |
Common optional (defaults in parentheses)
Variable | Description |
| Shared secret for the HTTP endpoint ( |
| Loan entity set name ( |
| Listen port/host ( |
|
|
See .env.example for the full list, including the WhatsApp settings below.
Usage
The server is transport-only; a client drives the tools.
Claude Desktop (stdio): point an
mcpServersentry atnode dist/index.js(ornpm run start:stdio).Copilot Studio / remote (HTTP): add an MCP tool at
https://<host>/mcpwith API-key auth (headerx-api-key).
GET /healthz returns liveness and WhatsApp status.
WhatsApp (optional)
Lets users query loans over WhatsApp. Enabled only when HUGGINGFACE_API_KEY and the TWILIO_* variables are set.
Set
HUGGINGFACE_API_KEY(or another OpenAI-compatible endpoint viaLLM_BASE_URL/LLM_MODEL).Create a Twilio WhatsApp Sandbox; set
TWILIO_ACCOUNT_SID,TWILIO_AUTH_TOKEN,TWILIO_WHATSAPP_FROM.In the sandbox settings, set the inbound webhook to
https://<host>/whatsapp(POST).Set
TWILIO_VALIDATE_SIGNATURE=trueonce the URL is stable.
Deployment
Deployed on Azure App Service. Pushing to main builds and deploys via GitHub Actions
(.github/workflows/main_loan-mcp-server.yml). Set all required and secret variables as
App settings; Azure runs npm start (the HTTP server) and injects PORT automatically.
Scripts
Script | Description |
| Compile TypeScript to |
| Run the HTTP server |
| Run the stdio server |
| Run from source with reload |
| Type-check without emitting |
| Integration test (requires a populated |
| Integration test over HTTP |
License
MIT
Available Tools
12 toolsExplainEligibilityExplain EligibilityB
Explain a loan application's eligibility in business-friendly language, including its eligibility status, remarks and whether manual review is required.
Inputs: referenceNumber (required).
GetApplicantHistoryGet Applicant HistoryA
Retrieve all previous loan applications for an applicant, identified by their email address, ordered by most recent first.
Inputs: applicantEmail (required).
GetLoanAnalyticsGet Loan AnalyticsA
Return portfolio-level analytics across all loan applications: totals by status and average/highest/lowest loan amounts.
Inputs: no named parameters.
GetLoanStatusGet Loan StatusA
Retrieve the current processing status of a single loan application by its reference number, including eligibility status, the assigned officer and whether manual review is required.
Inputs: referenceNumber (required).
GetLoanSummaryGet Loan SummaryA
Retrieve the full summary of a single loan application by its reference number, including applicant details, amounts, status, eligibility and the assigned officer.
Inputs: referenceNumber (required).
GetOfficerWorkloadGet Officer WorkloadA
Summarize a loan officer's workload: total assigned applications, pending reviews and a breakdown by status, with the underlying loan list.
Inputs: officerName (required).
GetRequiredDocumentsGet Required DocumentsA
Return the list of documents required for a given loan type. Falls back to a default checklist for unrecognized loan types.
Inputs: loanType (required).
ListPendingReviewsList Pending ReviewsA
List all loan applications currently in the manual-review queue (status = Under Review).
Inputs: no named parameters.
SearchLoansByOfficerSearch Loans By OfficerA
Find all loan applications assigned to a specific loan officer, identified by their full name.
Inputs: officerName (required).
SearchLoansByPhoneSearch Loans By PhoneA
Find all loan applications associated with a given phone number, returned as a compact list.
Inputs: phoneNumber (required).
SearchLoansByStatusSearch Loans By StatusA
Find all loan applications matching a given status label (e.g. Received, Pending, Under Review, Approved, Rejected).
Inputs: status (required).
TrackLoanTimelineTrack Loan TimelineA
Return a chronological timeline for a loan application using its created date, eligibility-check date, officer assignment and current status.
Inputs: referenceNumber (required).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 12 tools
Most tools have clear distinct purposes, such as GetLoanSummary, SearchLoansByPhone, and GetOfficerWorkload. However, a few overlap: GetLoanSummary and GetLoanStatus both return status and officer info, and TrackLoanTimeline provides timeline which overlaps with status. These overlaps could cause minor misselection.
All tool names follow a consistent Verb+Noun+Modifier pattern in PascalCase (e.g., GetLoanSummary, SearchLoansByStatus, ListPendingReviews). The naming is predictable and uniform across the set.
With 12 tools, the set is well-scoped for a loan management server. It covers retrieval, search, analytics, and officer workload without being excessive or insufficient.
The tools comprehensively cover retrieval queries (summary, status, timeline, search, analytics, documents) but lack any mutation operations (create, update, delete loans). This is a notable gap for a complete loan management surface.
Maintenance
Related MCP Connectors
MCP server for Codat — companies, connections, invoices, bills and financial statements.
Hosted MCP server for AI-driven data ops. Create apps, manage schemas, and CRUD structured data.
MCP server for the Seline Analytics API
An MCP server that provides read access to your cloud storage providers, bank accounts and more.
Related MCP Servers
- AlicenseNot gradedqualityBmaintenanceThe most complete MCP server for Microsoft Dataverse.271 npm18MIT
- AlicenseBqualityBmaintenanceMCP server for Microsoft Dataverse API with safe-by-default configuration. Works with any Dataverse / Dynamics 365 environment.23508 npm8MIT
- FlicenseNot gradedqualityDmaintenanceA production-grade MCP server that exposes real-time banking data replicated via Oracle GoldenGate CDC as structured tools for AI agents, enabling read, score, and write operations on customer, account, transaction, and alert data.1-
- FlicenseNot gradedqualityDmaintenanceA production-grade MCP server for a fictional digital bank, exposing tools for an AI copilot to service customers across the full risk spectrum from read-only lookups to money movement and destructive admin actions, with OAuth 2.1 security and a realistic dataset.23 npm1-