Latest Posts

Notes on AI agents, Copilot Studio, .NET and cloud engineering.

  1. Long-Running MCP Tools in C#: Call a Six-Second Tool as a Task and Poll It

    Call a six-second MCP tool as a task and poll it to completion with the C# Tasks extension: server and client code that ran, the client methods, and what to fix before production.

    By Ajith Joseph · Sat Nov 07 2026 · #dotnet #csharp #mcp #aspnet-core

  2. November 10: Moving .NET 8 and .NET 9 AI and MCP Projects to .NET 10

    .NET 8 and .NET 9 reach end of support on November 10, 2026. A checklist for moving AI and MCP projects to .NET 10, and why .NET 11 can wait in CI for now.

    By Ajith Joseph · Fri Nov 06 2026 · #dotnet #csharp #ai #aspnet-core

  3. Human Approval for Agent Tools with Microsoft Agent Framework in .NET

    Require a human yes before an agent runs a risky tool in Microsoft Agent Framework. Tested numbers for approve and reject, and the local-function naming trap that made it look broken.

    By Ajith Joseph · Thu Nov 05 2026 · #dotnet #csharp #security #ai-agents

  4. Copilot Studio Model Changes: The OpenAI Sub-Processor Switch and What to Check in Your Tenant

    GPT-6 Astra and GPT-5.6 Terra are available in Copilot Studio only when OpenAI is enabled as a sub-processor. A governance checklist for agent owners and tenant admins.

    By Ajith Joseph · Wed Nov 04 2026 · #security #copilot-studio #governance #ai-agents

  5. Testing .NET AI Apps with Microsoft.Extensions.AI.Evaluation

    Score model output in unit tests with Microsoft.Extensions.AI.Evaluation: the packages, a first relevance evaluation, how to assert safely, and rules for keeping an LLM judge repeatable.

    By Ajith Joseph · Tue Nov 03 2026 · #dotnet #csharp #ai #testing

  6. A Production Pipeline for IChatClient: Microsoft.Extensions.AI Middleware in Order

    A production pipeline for IChatClient with Microsoft.Extensions.AI 10.x: cache, function invocation, OpenTelemetry and logging, why the order matters, and two tests that prove it.

    By Ajith Joseph · Mon Nov 02 2026 · #dotnet #csharp #ai #middleware

  7. A Stateless .NET MCP Server for Copilot Studio: Build It, Prove It With curl, Then Connect It

    The smallest stateless ASP.NET Core MCP server Copilot Studio can use, how to prove it with curl before touching the portal, whether an older client still works, and a failure checklist.

    By Ajith Joseph · Sun Nov 01 2026 · #dotnet #copilot-studio #mcp #aspnet-core

  8. The MCP C# SDK 2.x Release Train: What Changed Since 2.0 and What Breaks When You Upgrade

    What changed between MCP C# SDK 2.0 and 2.3: stateless HTTP by default, Tasks as an extension, nullable ServerInfo and stricter OAuth, with an upgrade checklist tested against a real server.

    By Ajith Joseph · Sat Oct 31 2026 · #dotnet #csharp #mcp #aspnet-core

  9. Copilot Studio Custom Graders: Write the Rubric Once, Reuse It in Every Evaluation

    Copilot Studio is adding a reusable library of custom graders, generally available November 15, 2026. How to write graders that earn their place and keep the rubric under version control.

    By Ajith Joseph · Fri Oct 30 2026 · #copilot-studio #testing #ai-agents

  10. Copilot Chat Agents in Copilot Studio Reach GA: Topics, Connected Agents and a Real Lifecycle

    Microsoft says the enhanced Copilot Studio support for Copilot Chat agents reaches GA on October 26, 2026. What topics, connected agents, evaluations and lifecycle management change, and what to check first.

    By Ajith Joseph · Thu Oct 29 2026 · #copilot-studio #governance #ai-agents

  11. WebMCP Explained: Make Your Website Callable by AI Agents

    What WebMCP is, where it stands in October 2026, and how I added it to this blog: four tools, the exact code, real outputs, and the document.modelContext bug that hid them.

    By Ajith Joseph · Fri Oct 02 2026 · #web-development #mcp #webmcp #ai-agents #chrome

  12. Standard or GitHub Copilot Harness? A Decision Guide

    Copilot Studio now runs agents on three harnesses. What each is for, the real trade-offs in cost and control, and how to decide, and migrate, without a rewrite for its own sake.

    By Ajith Joseph · Thu Oct 01 2026 · #architecture #copilot-studio #github-copilot #ai-agents

  13. Copilot Studio Credits: Estimate, Cap and Avoid Surprises

    How Copilot Studio credits are billed in October 2026: the rate table, a tested calculator, the enforcement rules, and the controls that stop a prototype from becoming a bill.

    By Ajith Joseph · Thu Oct 01 2026 · #copilot-studio #cost-optimization #governance #copilot-credits

  14. Gate Copilot Studio Releases with the Evaluation API

    Run Copilot Studio agent evaluations from a pipeline with the Power Platform API, then block a release on regressions. A tested gate script, the auth caveats, and the pitfalls.

    By Ajith Joseph · Thu Oct 01 2026 · #copilot-studio #power-platform #testing #ci-cd #ai-agents

  15. Agents Need Identities, Not Shared API Keys: A Least-Privilege Playbook for 2026

    Survey data shows agent adoption outrunning security. How Entra Agent IDs, MCP auth changes, and a deny-by-default tool gate (tested C#) give AI agents least privilege.

    By Ajith Joseph · Tue Sep 29 2026 · #dotnet #security #authentication #copilot-studio #ai-agents

  16. EU AI Act After the Omnibus: What Moved, What Didn't, and a Developer Checklist

    The EU delayed high-risk AI Act rules to 2027 and 2028, but Article 50 transparency duties already apply. What changed, what didn't, and a practical checklist for developers.

    By Ajith Joseph · Tue Sep 29 2026 · #ai #eu-ai-act #compliance

  17. The September 2026 Price Ladder: Building a Cost-Aware Model Router

    Opus 5.5 and Sonnet 5.5 landed in September 2026. Verified Anthropic prices, the tokenizer trap, and a tested TypeScript router that cascades cheap to strong.

    By Ajith Joseph · Tue Sep 29 2026 · #typescript #ai #claude #llm #cost-optimization

  18. MCP's 2026-07-28 Spec: The Stateless Rewrite, and What Your Server Needs to Change

    The July 2026 MCP spec removes sessions and the initialize handshake. Verified changes, real requests against a C# server, gateway routing, and a migration checklist.

    By Ajith Joseph · Tue Sep 29 2026 · #dotnet #architecture #mcp #ai-agents

  19. Your Copilot Studio Agent's Model Will Change Under You: A Change-Management Playbook

    Copilot Studio upgrades default models and retires old ones. A tested playbook to pin, evaluate, compare runs, and diff exported results so a model change never surprises you.

    By Ajith Joseph · Tue Sep 29 2026 · #llm #copilot-studio #testing #ai-agents

  20. Putting a Copilot Studio Agent in Front of Customers: A Secure React Embed

    Embed a Copilot Studio agent in a React customer portal securely: a .NET token broker, opaque session handles, safe context passing, and tools that cannot be prompt-injected.

    By Ajith Joseph · Thu Sep 24 2026 · #react #dotnet #security #copilot-studio #ai-agents

  21. A Real Copilot Studio Agent: Wiring a .NET MCP Server to Your Order System

    A real-world Copilot Studio example with working C# code: build an MCP server over an existing order API, connect it as an agent tool, and ship it safely.

    By Ajith Joseph · Tue Sep 22 2026 · #dotnet #csharp #copilot-studio #mcp #ai-agents

  22. Building a Production Copilot Studio Agent in 2026: A Step-by-Step Guide

    A hands-on, step-by-step guide to building, grounding, governing, and shipping a Microsoft Copilot Studio agent on the 2026 GitHub Copilot harness.

    By Ajith Joseph · Tue Sep 22 2026 · #copilot-studio #power-platform #ai-agents

  23. Building an Agent Harness in .NET: Orchestrator and Sub-Agents with Dedicated Context

    Learn how to build a production agent harness in .NET, with an orchestrator, specialized sub-agents, dedicated context windows, and durable state.

    By Ajith Joseph · Sun Aug 02 2026 · #dotnet #llm #architecture #ai-agents

  24. Migrating from Claude Opus 4.7 to Opus 5: What Actually Changed for Agentic Coding

    A practical migration guide from Claude Opus 4.7 to Opus 5 for agentic coding workflows, covering cost math, prompt changes, and real gotchas.

    By Ajith Joseph · Sun Aug 02 2026 · #claude #llm #ai-agents

  25. Optimizing Core Web Vitals for AI-Heavy Next.js Applications

    Learn how to optimize Core Web Vitals for AI-heavy Next.js apps with actionable tips, tools, and strategies to improve LCP, FID, and CLS....

    By Ajith Joseph · Thu Jul 23 2026 · #ai #nextjs #web-performance