Recommended Free Tools
Keep your tool logic separate from the model provider, then pick the seam that matches what you want to reuse. If you want to swap models (Azure OpenAI, OpenAI, Ollama) without touching your tools, use Microsoft.Extensions.AI. If you want the same tools to work across different AI hosts and applications, expose them through an MCP server built with the official C# SDK. The two compose well. Neither one removes the work of implementing, authorizing and operating the tool itself.
What tool calling actually is
The model never runs your .NET method. It returns a structured request naming a tool and its arguments, your application invokes the tool, and the result goes back to the model, which continues until it can answer. Microsoft’s AI tool calling documentation for .NET states the safety consequence plainly: “Models might hallucinate arguments that weren’t described in your function definitions.” Validate arguments inside every tool, whichever integration route you take.
As an Amazon Associate I earn from qualifying purchases.
Two different portability problems
| Question | Seam to use | Where tools run |
|---|---|---|
| “I may change model providers.” | Microsoft.Extensions.AI function calling | In-process, inside your app |
| “Other hosts or apps should use the same tools.” | MCP server plus MCP clients | Separate server process or remote service |
| “Both.” | MCP client tools fed into Microsoft.Extensions.AI | Mixed |
Option 1: Microsoft.Extensions.AI for provider independence
Microsoft.Extensions.AI abstracts .NET interactions with AI providers. Per the Microsoft Learn guide to .NET AI and MCP, its function-calling pieces are:
Free tools Windows power users keep installed
One-click scans. No signup required.
AIFunction: a described function the model can request.AIFunctionFactory: wraps an existing .NET method as anAIFunction, so you don’t rewrite the method.FunctionInvokingChatClient: automates the loop of invoking requested functions, returning results and continuing until a final answer.
Microsoft lists Azure OpenAI, OpenAI and Ollama among the implementations. Your tool methods stay the same when you change the underlying chat client. Capabilities still vary by model and provider, so check that your chosen model supports function calling at all.
#1 Best Overall
Keep the tools provider-neutral
- Write tools as plain methods with clear parameter names and descriptions, with no provider SDK types in signatures.
- Register them through
AIFunctionFactoryat the edge of your app. - Put authorization and argument validation inside the tool, not in the prompt.
Mind the token cost
Tool definitions count toward the model’s token limit. Microsoft suggests registering fewer tools, shortening names and descriptions, and offering only tools relevant to the conversation. Parallel calls are handled by FunctionInvokingChatClient only when the underlying model supports them, so check your provider’s documentation rather than assuming it.
Option 2: MCP for reuse across hosts
The Model Context Protocol standardizes a different boundary. Its architecture has a host, MCP clients and MCP servers, and clients can list and call a server’s tools. The official MCP C# SDK supports building both .NET clients and servers. A tool exposed once as an MCP server can be discovered by any compliant host, and your own assistant can consume it as a client and hand the discovered tools to the model through Microsoft.Extensions.AI.
Package roles
The SDK repository describes these packages:
| Package | Use it for |
|---|---|
ModelContextProtocol.Core |
Clients and low-level APIs |
ModelContextProtocol |
Most servers, with hosting and dependency injection |
ModelContextProtocol.AspNetCore |
HTTP-based MCP servers |
| Apps and Tasks extensions | Separate, more specialized packages |
The repository is rolling documentation, so confirm current package versions and install guidance there before copying commands from any article.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Version-sensitive details
Microsoft’s announcement of v2.0 of the SDK describes stateless behavior as the default and says v2 prefers protocol revision 2026-07-28, while retaining down-level behavior in stated cases. It lists targets of net8.0, net9.0, net10.0 and netstandard2.0, and calls out migration differences for anyone using the experimental Tasks feature. In the post, Microsoft Engineering Manager Jeff Handley wrote: “The 2.0 release of the MCP C# SDK is a milestone for building MCP servers and clients on .NET.” These details were current as of this article’s October 2026 writing. Read the release notes for the version you install.
What MCP costs you
MCP adds a protocol, a transport choice, server lifecycle management and authorization decisions. Local function calling keeps execution in your process, with less to run. MCP does not mean “write once, works everywhere”: tool semantics, permissions, provider capabilities and server operations still need deliberate design.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How to choose
- One app, may switch models: use Microsoft.Extensions.AI and
AIFunctionFactoryonly. - Several apps or third-party hosts need the same tools: wrap the tools in an MCP server, using the ASP.NET Core package for HTTP.
- Your assistant must also use tools others publish: add an MCP client and pass the discovered tools to your chat client.
- Many tools: filter by conversation relevance to limit token use.
A sensible path is to start in-process. Because the tools are plain methods, moving them behind an MCP server later is a packaging change, not a rewrite.
Quick Recap
Rank #4
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
