Insights, AI Innovation & Technology Leadership

At MagnusMinds, we strongly believe that successful collaboration goes beyond virtual meetings and emails. As our organization continues to grow, our senior team members are actively traveling to client locations to better understand business requirements, streamline processes, and ensure seamless project execution. These on-site engagements help us build stronger relationships, improve communication, and deliver solutions more effectively. Strengthening Global Partnerships Through On-Site Collaboration UK Visit Earlier this year, our CEO/Founder visited one of our valued clients in the United Kingdom to discuss long-term technology strategies, operational improvements, and future collaboration opportunities. The visit focused on understanding evolving business requirements, aligning technical processes, and ensuring smooth execution of ongoing initiatives. Such leadership-level interactions help us create a stronger foundation for long-term partnerships. Dubai Visit Recently, one of our Project Managers travelled to Dubai for 4 weeks to work closely with a client on a specialized .NET and Gaming Integration project. The visit involved technical discussions, architecture planning, integration assessments, and collaborative development workshops. Being on-site allowed our team to gain a deeper understanding of the client’s expectations and accelerate project progress efficiently. Pune Visit One of our Team Leads visited Pune as part of an ongoing engagement with an existing client. The primary objective of the visit was to review project milestones, optimize workflows, and ensure seamless coordination between teams. Face-to-face collaboration enabled faster decision-making and strengthened the overall execution process. Gurgaon-Delhi Visit As part of our commitment to delivering structured and scalable solutions, one of our Data Architect recently visited Delhi to establish processes for an upcoming project. The visit focused on requirement gathering, workflow planning, team alignment, and defining delivery frameworks to ensure a smooth project kickoff and successful implementation. Mumbai Visits Our commitment to client success is reflected in the continuous efforts of our team members who regularly visit a client’s office in Mumbai almost every month. These recurring visits help maintain strong communication, monitor project progress, address challenges proactively, and ensure that collaboration remains efficient and productive. Growing Together with Our Clients These visits are a reflection of how MagnusMinds is continuously evolving as a trusted technology partner. We believe that direct collaboration, proactive communication, and on-site engagement create better outcomes for every project we undertake. If your organization is looking for a dedicated technology partner, our team would be happy to collaborate closely with you, including visiting your workplace whenever needed to ensure project success, seamless communication, and long-term value creation. At MagnusMinds, we don’t just deliver solutions, we build partnerships. Let’s build something impactful together!

Why MCP servers?  LLMs are powerful—but they’re limited to what they can ‘see’. The Model Context Protocol (MCP) is an open protocol that standardizes how apps expose tools, resources, and prompts to AI clients so models can interact with real systems in a structured, discoverable way.  For .NET developers, this is especially useful because you can build MCP servers in C# using the official MCP C# SDK and run servers locally over stdio or remotely over HTTP.  MCP mental model (fast)  Host: The application that contains the AI experience (IDE/agent tool).  Client: The MCP-capable component inside the host that connects to servers.  Server: Your service that exposes tools/resources/prompts.  Choosing a transport: STDIO vs Streamable HTTP  STDIO (local): The client launches your server as a subprocess and communicates via stdin/stdout. Messages are newline-delimited JSON-RPC, and stdout must contain only protocol messages (logs must go to stderr).  Streamable HTTP (remote/scalable): Runs as an independent server reachable over HTTP. Validate the Origin header to reduce DNS rebinding risk and bind to localhost for local runs.  Part 1 — The fastest way: .NET 10 MCP Server Project Template  Microsoft provides a quickstart showing how to create a minimal MCP server using the .NET 10 SDK and the Microsoft.McpServer.ProjectTemplates template package.  This path is great for getting a working server quickly with correct wiring and sane defaults.  dotnet new install Microsoft.McpServer.ProjectTemplates Part 2 — Build a Minimal STDIO MCP Server (from scratch)  STDIO is ideal when your MCP server needs access to local machine resources and you want the simplest deployment path.  Below is a minimal server that uses Microsoft.Extensions.Hosting and exposes one tool (Echo).    Step A — Create project & add packages dotnet new console -n MyMcpServer cd MyMcpServer dotnet add package ModelContextProtocol dotnet add package Microsoft.Extensions.Hosting Note: Microsoft’s MCP server walkthrough shows the SDK approach using Microsoft.Extensions.Hosting and MCP server registration in the builder Step B — Program.cs (STDIO server + tool discovery) using Microsoft.Extensions.Hosting; using Microsoft.Extensions.DependencyInjection; using Microsoft.Extensions.Logging; using ModelContextProtocol.Server; using System.ComponentModel; var builder = Host.CreateApplicationBuilder(args); // IMPORTANT: STDIO servers must keep stdout clean. // Route logs to stderr so they don't corrupt JSON-RPC output. builder.Logging.AddConsole(o => o.LogToStandardErrorThreshold = LogLevel.Trace); builder.Services .AddMcpServer() .WithStdioServerTransport() .WithToolsFromAssembly(); await builder.Build().RunAsync(); [McpServerToolType] public static class EchoTools { [McpServerTool, Description("Echoes the message back to the client.")] public static string Echo([Description("Message to echo")] string message) => $"Hello from MCP (.NET 10): {message}"; } Why the stderr logging rule matters The MCP transport spec explicitly requires that in stdio, servers must not write non-protocol output to stdout; logs should go to stderr   Part 3 — Build a Remote MCP Server over Streamable HTTP (ASP.NET Core)  If you want a centrally hosted MCP server (team-wide tooling, enterprise integrations), use HTTP transport. MCP’s spec notes Streamable HTTP is the standard remote transport and includes security requirements like Origin validation. [modelconte...rotocol.io], [dometrain.com]Below is a minimal HTTP server exposing a demo Weather tool.  Step A — Create ASP.NET Core app & add MCP server support   dotnet new web -n MyHttpMcpServer cd MyHttpMcpServer dotnet add package ModelContextProtocol.AspNetCore Step B — Program.cs (HTTP MCP endpoint) using ModelContextProtocol.Server; using System.ComponentModel; var builder = WebApplication.CreateBuilder(args); builder.Services .AddMcpServer() .WithHttpTransport(options => { // Stateless mode is commonly recommended for simple remote servers // that don't need advanced server->client features. options.Stateless = true; }) .WithToolsFromAssembly(); var app = builder.Build(); // Exposes /mcp endpoint (or the configured MCP endpoint) app.MapMcp(); app.Run("http://localhost:3001"); [McpServerToolType] public static class WeatherTools { [McpServerTool, Description("Returns a sample weather status for a city.")] public static string GetWeather(string city) => $"Weather for {city}: Sunny (demo)"; } Security note (important): For Streamable HTTP, MCP recommends validating the Origin header to prevent DNS rebinding and binding locally to localhost when running locally   Part 4 — Coding standards & best practices for MCP servers  STDIO rule: stdout must be pure JSON-RPC Never log to stdout. Use stderr.  Standard: Configure logging to stderr as shown in the STDIO sample. Keep tools small + deterministic Tool methods should be short, validate input, and return structured outputs. Avoid tools that do “too much” (hard to reason about / secure). Validate inputs like public APIs Even though an LLM is “calling” the tool, treat it like an untrusted client: Validate required fields Constrain sizes Apply allowlists where possible Prefer “read-only” tools first Start with: search/read/query tools Then move to “write” tools with extra safety checks. Remote servers must follow transport security guidance Streamable HTTP transport includes security requirements like Origin validation to mitigate DNS rebinding risks Conclusion  .NET 10 makes it practical to build MCP servers using local STDIO transport for quick, secure local tooling, and Streamable HTTP for scalable, shared integrations.  Start with a small set of safe tools, add observability and security early, and expand capabilities over time. 

A professional services client had a clear idea: finance teams and business stakeholders spend too much time finding answers in Xero instead of acting on them. We took ownership of the solution end-to-end - design, development, security model, and go-to-market - delivering a Finance Agent for Xero that works inside Microsoft Teams and can also be used as an extension point for Microsoft 365 Copilot experiences, so users can ask questions where they already work. The challenge (pain points in real operations) As the client scaled, finance operations didn’t just get “busier”—they got noisier and slower, because many business-critical answers lived behind manual steps and finance-team dependency: Hours lost in coordination: Business users (project managers, sales ops, leadership) needed frequent answers—“Which bills are pending?”, “What’s overdue?”, “Did this invoice go out?”, “What’s the latest P&L?”. The only reliable path was messaging the finance team, waiting for someone to check Xero, clarifying filters, and repeating the loop—often consuming hours end-to-end for what should be a quick question. Manual filtering and inconsistent results: Invoices and bills were searched by multiple criteria (date ranges, contacts, status, amounts, references). Different people used different filter combinations, so results could vary, creating rework and follow-up questions. Draft categorization bottleneck: Draft bills required coding/categorization before review and approval. During peak periods, drafts piled up, approvals slowed down, and finance had to spend time fixing categorization inconsistencies. Reporting delays for stakeholders: Leadership wanted near-instant Profit & Loss visibility and “attainment-style” performance views, but reports still required manual pull, formatting, and explanation—creating delays in meetings and decision cycles. Adoption risk from sign-in friction: Any solution that forces frequent reconnects fails in practice. Xero OAuth also has real constraints (refresh tokens can expire if unused, and refresh responses can include a new refresh token that must be stored), so token handling had to be designed for reliability.?   What we built (tailored solution, delivered end-to-end) We designed the agent around actual finance requests and approval behaviors, not around technical endpoints. 1) Secure tenant connection with uninterrupted access Users connect their Xero tenant with OAuth 2.0, which allows an app to access Xero data via permission scopes after user approval (without needing the user’s password).? Because Xero access tokens expire quickly (30 minutes), we designed the agent to refresh access automatically to keep the Teams/Copilot experience uninterrupted.   2) Natural-language finance operations (self-serve answers) Inside Teams (and surfaced through Copilot extensibility patterns), users can ask in plain language and get results in minutes: Invoices & bills retrieval using conversational filters (who, when, status, amount, references). Profit & Loss on demand using Xero’s reporting endpoints (including Profit & Loss support), so stakeholders can pull the view they need without waiting on finance.? Profit and revenue attainment views based on the organization’s tracking logic—delivering “decision-ready” outputs rather than raw tables.   3) Automated draft categorization with human review To remove the biggest operational choke point while keeping governance intact: When a bill is in Draft, the agent applies predefined categorization logic and proposes the coding. It then posts the recommendation back to chat for approve/reject, so finance retains control while eliminating repetitive preparation work.   Real-world impact (what changed after rollout) From hours to minutes: Before the agent, users often spent hours coordinating with finance to get invoice/bill answers and report snapshots. After rollout, many of those requests were completed in minutes through the agent in Teams—reducing finance dependency and speeding decisions. Efficiency gains across the finance workflow: Finance teams spent less time on repetitive lookups and pre-approval preparation, and more time on review, exceptions, and higher-value analysis. Improved consistency and reduced rework: Standardized categorization recommendations plus the approve/reject step reduced variability in coding and cut down on “fix it later” cleanups. Higher satisfaction and adoption: A consistent “ask in Teams / Copilot, get an answer” experience improved trust and usage, while explicit consent ensured the access model stayed enterprise-friendly.? Value beyond one tenant: Publishing to the Teams Store made these workflow improvements accessible to a broader audience of Xero users facing the same operational friction.

Many organizations use Xero as their primary accounting system and rely on its built-in reports for financial review. While these reports are accurate and reliable, they are designed for basic financial visibility, not for deeper analytics or visual exploration.  At MagnusMinds, we worked with a client who wanted to move beyond traditional accounting reports and gain clear, visual, decision-ready insights from their financial data without disrupting existing accounting processes.  The Reporting Limitation We Identified  The client was already using Xero effectively for accounting. However, their reporting workflow revealed a familiar pattern:  Financial reports were reviewed in tabular format  Limited visualization made trend analysis difficult  Comparing performance across periods required manual effort  Insights depended heavily on interpretation rather than visuals  The challenge wasn’t data accuracy it was how that data was being consumed.  Our Approach: Extending, Not Replacing Xero  Rather than extracting raw accounting transactions and recreating financial logic externally, we designed a solution that respected Xero as the system of record.  Our focus was on extending Xero’s reporting, not rebuilding it.  What We Did  1. Leveraged Xero’s Native Reports  We identified Xero’s financial reporting APIs as the most reliable source for analytics-ready data. This allowed us to work with figures that already aligned with Xero’s Profit & Loss and other financial statements.  2. Built a Custom API Integration  We implemented a custom API layer to extract financial report data from Xero automatically. This eliminated the need for manual exports and ensured consistent, repeatable data retrieval.  3. Structured the Data for Analytics  The extracted data was transformed into a clean, structured format suitable for Power BI. We kept the model intentionally lightweight to avoid unnecessary complexity.  4. Enabled Advanced Visualisation in Power BI  With accurate financial data available, we built interactive Power BI dashboards that introduced:  Profit & Loss visualisations with monthly and quarterly trends Revenue and expense breakdowns by account category and reporting period Comparative Profit & Loss views across financial periods Net profit and margin analysis with visual indicators Operating expense analysis aligned to Xero chart of accounts Period-over-period variance analysis for income and expenses Executive summary dashboards reflecting Xero’s financial statements at a glance All without altering the underlying accounting logic.  The Impact of Our Work  The solution delivered immediate and measurable benefits:  Financial reports became visually intuitive and easier to interpret  Manual reporting effort was significantly reduced  Data consistency with Xero was preserved  Leadership gained faster access to actionable insights  Reporting scaled effortlessly as business needs evolved  Most importantly, the client moved from reviewing numbers to understanding performance.  Why This Approach Works  Accounting systems and analytics platforms serve different purposes. By clearly separating responsibilities Xero for accounting and Power BI for analytics we avoided unnecessary risk and complexity.  Our approach ensured:  Accuracy was never compromised  Reporting remained flexible and scalable  Analytics evolved without impacting accounting operations  From Accounting to Analytics  This demonstrates how organisations can unlock greater value from existing accounting systems. With the right integration strategy, basic financial reports can be transformed into powerful analytical assets.  At MagnusMinds, we help organisations bridge the gap from accounting to analytics, turning trusted financial data into meaningful business insight.   

Measure Killer: The Power BI External Tool We Wish We Had Found Sooner We'll be honest. The first time someone on our team heard "Measure Killer," the reaction was: wait, you want us to install that on a client's report? But then at MagnusMinds we tried it. Now it's one of the first tools we open on any Power BI model older than a few months. This quick guide shows exactly what Measure Killer does, why it matters, and how it keeps your reports clean, fast, and easy to hand over. Key Concepts What Measure Killer Actually Does Measure Killer is a free external tool for Power BI Desktop built by Kurt Buhler. It scans your open PBIX file and finds every unused measure, column, and table. It understands full DAX dependency chains so a measure called by another active measure is never flagged as unused. No setup, no subscription, no cloud just click from the External Tools ribbon and it works instantly. It shows the DAX expression right in the tool so you can review before deleting anything.                      Practical Use Cases How We Use It Every Day Open any PBIX → External Tools → click Measure Killer → scan takes seconds. Review the list of unused items (we usually see 20–30 % of measures are dead weight). Keep anything still in progress; select the rest and delete in one click. Works on leftover tables, old date tables, renamed columns everything that bloats your model. Run it before handing over to clients, before production, or when a file feels “slow.” Key Benefits – Why It Belongs in Every Toolkit Faster reports – smaller models refresh quicker and load faster for users. Easier maintenance – no more hunting through 130+ measures to find what’s active. Cleaner handovers – new developers instantly see only what matters. Completely safe – you stay in full control; nothing deletes without your confirmation. Free forever – no catch, no paid tier. Turns months of hidden clutter into a 15-minute cleanup job. Quick Example A client’s finance dashboard had 132 measures. After one Measure Killer scan we removed 78 unused ones. File size dropped 40 MB and refresh time halved. Total time: 12 minutes. Conclusion Measure Killer doesn’t promise magic, it just quietly solves a real problem that every Power BI developer faces. Download it, run it on any report you haven’t cleaned lately, and you’ll see the difference in seconds. Your models will thank you and so will your team. Ready to try it? Grab the free tool from the Power BI community and make it part of your standard process today.  

Introduction In modern AI applications, Retrieval-Augmented Generation (RAG) is one of the most powerful patterns for building intelligent assistants that go beyond generic responses. Instead of relying only on pre-trained knowledge, RAG allows your assistant to retrieve real data (files, database, CSV, APIs) and generate accurate, context-aware answers. In this blog, we will walk through how to build an OpenAI Assistants Bot using RAG, step by step.   What is RAG? RAG (Retrieval-Augmented Generation) is a technique where: Data is stored (files, DB, vector store) User asks a question Relevant data is retrieved AI generates response based on retrieved data Instead of hallucinating, AI gives real, data-backed answers Step 1 — Register on OpenAI Platform Open: https://platform.openai.com/ Important Notes: Create an account using your email If using for organization: Ask admin to add you to organization OpenAI provides: Personal account Organization account   Step 2 — Generate API Key Go to: Settings → API Keys Important: You need to manage two types of keys properly Type Usage Personal API Key For personal development Organization API Key For production / team usage Best Practice Never hardcode API keys Store in: .env file or Azure Key Vault `` Step 3 — Open Playground & Assistant Go to: OpenAI Dashboard → Playground → Assistants First time? It will prompt: Create a new Assistant OpenAI auto-generates: A random name You can rename it (recommended) What is an Assistant? An Assistant is: A configured AI agent With instructions With connected tools With uploaded data (RAG-ready) It acts as a "brain" that connects: Instructions Files APIs Conversations Step 4 — Add Data (Core of RAG) Example You are an AI assistant for an e-commerce store. Answer user questions based only on the provided data. Do not generate fake information. This is very important for RAG behavior. Now comes the most important part — giving data to your assistant. Supported Data Types PDF CSV TXT DOCX JSON Example: E-Commerce Data (CSV) If you have an e-commerce website, create CSV like: ProductId,Name,Category,Price,Stock 1,iPhone 15,Mobile,999,20 2,Samsung S23,Mobile,899,15 3,MacBook Pro,Laptop,1999,10 `` Upload Process Go to Assistant Open Files section Upload CSV / document Attach file to assistant How RAG Works in Assistants https://via.placeholder.com/1200x500?text=Assistant+RAG+Flow Flow: User asks:  “What is price of iPhone 15?” Assistant: Searches uploaded data Retrieves relevant content AI generates answer: “The price of iPhone 15 is $999” Step 5 — Call Assistant via API Example (C# style for you ): var client = new OpenAIClient("API_KEY"); var thread = await client.CreateThreadAsync(); await client.AddMessageAsync(thread.Id, "What is the price of iPhone 15?"); var response = await client.RunAssistantAsync(thread.Id, assistantId); Console.WriteLine(response); Conclusion OpenAI Assistants + RAG is one of the most powerful ways to build: Intelligent Context-aware Production-ready AI applications Instead of building complex pipelines, you can now: Upload data Configure assistant Start querying