Build MCP is a free Model Context Protocol server that connects your AI coding assistant directly to the Avalonia documentation. It's available now, it works with every major AI-enabled editor, and it fundamentally changes the quality of AI-generated Avalonia code.
If you've used an AI assistant with Avalonia before, you'll know the problem. The answers sound confident but they're often wrong: WPF syntax mistaken for Avalonia, properties that don't exist, APIs from several versions ago. The model is working from whatever scraps ended up in its training data, and for a framework that moves as quickly as ours, that's not good enough. Build MCP fixes this by giving your AI assistant real-time access to the guides, tutorials, and API references that actually reflect how Avalonia works today.
The Biggest Documentation Update We've Ever Shipped
Before we get into the tooling, I want to share some context. With Avalonia 12, we've completely overhauled our documentation. There are now over 125% more guides than before, covering everything from getting started to advanced topics that previously lived only in GitHub issues or community tribal knowledge. The API reference has been fully refreshed too, so every class, property, method, and event is documented against the current codebase.
This is important because Build MCP draws from this content directly. The more comprehensive and accurate the docs, the better your AI assistant performs.
The new v12 documentation portal is live now at docs.avaloniaui.net, and it includes a built-in Ask AI feature powered by GPT-4.1 Mini for quick lookups while you're browsing. It's handy for a fast answer when you're already on the site. But if you want the best experience, where your AI assistant has the full power of its reasoning, your local project context, and the Avalonia documentation all working together, Build MCP is what you want to be using.
What Exactly Is Build MCP?
If you haven't encountered MCP yet, here's the short version: Model Context Protocol is an open standard that lets AI models call external tools and services through a structured interface. Instead of you being the middleman (copy this, paste that, no not that bit, the other bit), your AI assistant talks to our documentation directly.
Build MCP exposes five tools to your assistant:
search_avalonia_docs: Searches the full Avalonia documentation with smart routing that optimises queries based on whether you're asking about styling, binding, MVVM patterns, or anything else.lookup_avalonia_api: Targeted lookups for specific classes, properties, methods, and events against the current API reference.get_avalonia_expert_rules: Loads a comprehensive set of development rules covering AXAML syntax, the property system, theming, threading, and common mistakes. Call it at the start of a session and your assistant gets a crash course in writing idiomatic Avalonia code.migrate_diagnostics: Guides you through upgrading to the current Developer Tools package, including package replacement and code updates.migrate_to_xpf: Step-by-step instructions for converting WPF applications to Avalonia using XPF, covering NuGet configuration and licence setup.
It also ships with four prompts that configure your assistant for specific workflows:
init: Sets up an expert Avalonia session for an existing project by loading development rules and configuring the assistant to use the documentation tools.new: Walks you through creating a fresh Avalonia application, from template selection to developer tools installation.wpf-migration: Analyses your WPF project and recommends the right migration path, automatically calling the appropriate migration tools.recreate-ui: This is the one I want to show you.
From Screenshot to Running App
I wanted to see how far the recreate-ui workflow could go, so I ran an experiment. I grabbed a screenshot of an application I liked the look of, fed it to Claude Code with Build MCP connect, and typed: "recreate this app with Avalonia." That was the entire prompt. No AXAML skeleton. No design spec. Just the image and a single sentence.
Claude studied the screenshot, loaded the Avalonia expert rules through Build MCP, and started writing AXAML. Using DevTools MCP, it then attached to the running app, captured a screenshot of what it had produced, compared it to the original, and started iterating. It adjusted spacing, refined colours, reworked the layout hierarchy, and kept going until the result matched. The whole process took minutes, not hours.
The result is a fully functional Avalonia application, running natively on desktop, built entirely from a single screenshot with zero manual intervention. Not a pixel-perfect clone, but a faithful recreation that captures the design intent and runs on Windows, macOS, and Linux.
This is the kind of workflow that changes how you think about prototyping. See something you like? Screenshot it. Let your AI assistant turn it into a working Avalonia app while you get on with the interesting problems. Come back, review what it's produced, iterate on the bits that need your judgment. The tedious translation from "I want it to look like this" to running code is handled for you.
Always Current, Always Accurate
AI assistants are only as good as the context they have. A general-purpose LLM knows a bit about a lot of frameworks, but the nuances of any specific one tend to get lost in the noise. Avalonia has its own conventions, its own property system, its own approach to styling and theming. These aren't the same as WPF's, and they're certainly not the same as what a model might have absorbed from a handful of Stack Overflow answers circa 2022.
Build MCP solves this by connecting your assistant to the documentation in real time. When you ask about data binding in a TreeView, it searches the actual docs. When you need to understand how StyledProperty differs from a dependency property, it looks it up. The difference in output quality is substantial.
Crucially, this isn't a static snapshot that you need to worry about keeping up to date. We rebuild its remote index from our documentation at least once a day. When we update a guide, fix an example, or document a new API, every agent connected to Build MCP picks up those changes without you lifting a finger. The knowledge stays current so you don't have to think about it.
Part of a Bigger Picture
Build MCP is free and stands on its own, but it's also the foundation of something more ambitious. We've built three MCP servers that together cover the full Avalonia development lifecycle.
DevTools MCP gives your AI assistant eyes and hands inside your running application. It can attach to a live Avalonia app or the XAML previewer, inspect the visual tree, search for elements by type or name, read and modify properties at runtime, capture screenshots, and send input events. Instead of describing a layout bug in words and hoping your assistant understands, it can see the problem, inspect the relevant properties, and suggest or apply a fix. This is the server that powers the iterative screenshot comparison in the recreate-ui workflow I described above.
Parcel MCP handles packaging and distribution. Your assistant can create Parcel configurations from your .NET projects, set up code signing and notarisation for macOS, configure Azure Trusted Signing for Windows, and build installers for every platform. Describe what you want in plain English and the AI handles the configuration. No more wrestling with signing certificates at 11pm before a release.
Together, these three servers cover build, debug, and ship. That's not a chatbot bolted onto your IDE. That's an AI-powered workflow across the full development lifecycle.
DevTools MCP and Parcel MCP are part of Avalonia Accelerate and require a licence. Build MCP is completely free, no licence key, no sign-up, no usage limits, because we believe that every developer building with Avalonia deserves an AI assistant that actually knows the framework.
Works Where You Work
We've built setup support for every major AI-enabled editor and CLI tool: VS Code with GitHub Copilot, Visual Studio, JetBrains Rider, Cursor, Windsurf, Claude Code, Claude Desktop, and Gemini CLI all have documented configuration paths. In most cases it's a single JSON snippet or a one-line terminal command.
For VS Code, you can add it through the command palette (search for "MCP: Add Server," choose HTTP, paste the URL) or drop a config file into .vscode/mcp.json. Rider users can go through the AI Assistant settings or create an .idea/mcp.json. Claude Code users get the most satisfying setup: claude mcp add --transport http avalonia-docs https://docs-mcp.avaloniaui.net/mcp. One command. Done. Go build something.
A Note on AI and Frameworks
You might have noticed that we don't lead with "AI-ready!" on our homepage or plaster it across every piece of marketing material. That's deliberate.
AI tooling support is something we consider table stakes. Every serious framework should work well with AI assistants, and we don't typically think of it as a feature worth trumpeting. We've invested heavily in Build MCP, DevTools MCP, and Parcel MCP because they genuinely make developers more productive.
The reason people choose Avalonia hasn't changed. It's the rendering engine. It's genuine cross-platform reach across desktop, mobile, web, and embedded. It's the depth of the control library and the styling system. It's the fact that companies like JetBrains, Unity, and Devolutions trust it for production applications used by millions. AI tooling makes all of that easier to work with, and that's genuinely valuable, but it's one feature in a much more compelling story.
We'd rather be known for building a great framework that happens to have excellent AI tooling than for building AI tooling that happens to sit on top of a framework. The substance comes first. Everything else follows from that.
Get Started
Head over to the Build MCP documentation for step-by-step setup instructions for your editor of choice. You'll be up and running in under a minute, and your AI assistant will finally know what it's talking about when it comes to Avalonia.
If you're already an Accelerate subscriber, set up DevTools MCP and Parcel MCP while you're at it. The three servers together give your assistant the ability to learn the framework, see your application, and ship it. Try the recreate-ui prompt with a screenshot of something you'd like to build. I'm genuinely curious to see what you create.
Build MCP is a free Model Context Protocol server that connects your AI coding assistant directly to the Avalonia documentation. It's available now, it works with every major AI-enabled editor, and it fundamentally changes the quality of AI-generated Avalonia code.
If you've used an AI assistant with Avalonia before, you'll know the problem. The answers sound confident but they're often wrong: WPF syntax mistaken for Avalonia, properties that don't exist, APIs from several versions ago. The model is working from whatever scraps ended up in its training data, and for a framework that moves as quickly as ours, that's not good enough. Build MCP fixes this by giving your AI assistant real-time access to the guides, tutorials, and API references that actually reflect how Avalonia works today.
The Biggest Documentation Update We've Ever Shipped
Before we get into the tooling, I want to share some context. With Avalonia 12, we've completely overhauled our documentation. There are now over 125% more guides than before, covering everything from getting started to advanced topics that previously lived only in GitHub issues or community tribal knowledge. The API reference has been fully refreshed too, so every class, property, method, and event is documented against the current codebase.
This is important because Build MCP draws from this content directly. The more comprehensive and accurate the docs, the better your AI assistant performs.
The new v12 documentation portal is live now at docs.avaloniaui.net, and it includes a built-in Ask AI feature powered by GPT-4.1 Mini for quick lookups while you're browsing. It's handy for a fast answer when you're already on the site. But if you want the best experience, where your AI assistant has the full power of its reasoning, your local project context, and the Avalonia documentation all working together, Build MCP is what you want to be using.
What Exactly Is Build MCP?
If you haven't encountered MCP yet, here's the short version: Model Context Protocol is an open standard that lets AI models call external tools and services through a structured interface. Instead of you being the middleman (copy this, paste that, no not that bit, the other bit), your AI assistant talks to our documentation directly.
Build MCP exposes five tools to your assistant:
search_avalonia_docs: Searches the full Avalonia documentation with smart routing that optimises queries based on whether you're asking about styling, binding, MVVM patterns, or anything else.lookup_avalonia_api: Targeted lookups for specific classes, properties, methods, and events against the current API reference.get_avalonia_expert_rules: Loads a comprehensive set of development rules covering AXAML syntax, the property system, theming, threading, and common mistakes. Call it at the start of a session and your assistant gets a crash course in writing idiomatic Avalonia code.migrate_diagnostics: Guides you through upgrading to the current Developer Tools package, including package replacement and code updates.migrate_to_xpf: Step-by-step instructions for converting WPF applications to Avalonia using XPF, covering NuGet configuration and licence setup.
It also ships with four prompts that configure your assistant for specific workflows:
init: Sets up an expert Avalonia session for an existing project by loading development rules and configuring the assistant to use the documentation tools.new: Walks you through creating a fresh Avalonia application, from template selection to developer tools installation.wpf-migration: Analyses your WPF project and recommends the right migration path, automatically calling the appropriate migration tools.recreate-ui: This is the one I want to show you.
From Screenshot to Running App
I wanted to see how far the recreate-ui workflow could go, so I ran an experiment. I grabbed a screenshot of an application I liked the look of, fed it to Claude Code with Build MCP connect, and typed: "recreate this app with Avalonia." That was the entire prompt. No AXAML skeleton. No design spec. Just the image and a single sentence.
Claude studied the screenshot, loaded the Avalonia expert rules through Build MCP, and started writing AXAML. Using DevTools MCP, it then attached to the running app, captured a screenshot of what it had produced, compared it to the original, and started iterating. It adjusted spacing, refined colours, reworked the layout hierarchy, and kept going until the result matched. The whole process took minutes, not hours.
The result is a fully functional Avalonia application, running natively on desktop, built entirely from a single screenshot with zero manual intervention. Not a pixel-perfect clone, but a faithful recreation that captures the design intent and runs on Windows, macOS, and Linux.
This is the kind of workflow that changes how you think about prototyping. See something you like? Screenshot it. Let your AI assistant turn it into a working Avalonia app while you get on with the interesting problems. Come back, review what it's produced, iterate on the bits that need your judgment. The tedious translation from "I want it to look like this" to running code is handled for you.
Always Current, Always Accurate
AI assistants are only as good as the context they have. A general-purpose LLM knows a bit about a lot of frameworks, but the nuances of any specific one tend to get lost in the noise. Avalonia has its own conventions, its own property system, its own approach to styling and theming. These aren't the same as WPF's, and they're certainly not the same as what a model might have absorbed from a handful of Stack Overflow answers circa 2022.
Build MCP solves this by connecting your assistant to the documentation in real time. When you ask about data binding in a TreeView, it searches the actual docs. When you need to understand how StyledProperty differs from a dependency property, it looks it up. The difference in output quality is substantial.
Crucially, this isn't a static snapshot that you need to worry about keeping up to date. We rebuild its remote index from our documentation at least once a day. When we update a guide, fix an example, or document a new API, every agent connected to Build MCP picks up those changes without you lifting a finger. The knowledge stays current so you don't have to think about it.
Part of a Bigger Picture
Build MCP is free and stands on its own, but it's also the foundation of something more ambitious. We've built three MCP servers that together cover the full Avalonia development lifecycle.
DevTools MCP gives your AI assistant eyes and hands inside your running application. It can attach to a live Avalonia app or the XAML previewer, inspect the visual tree, search for elements by type or name, read and modify properties at runtime, capture screenshots, and send input events. Instead of describing a layout bug in words and hoping your assistant understands, it can see the problem, inspect the relevant properties, and suggest or apply a fix. This is the server that powers the iterative screenshot comparison in the recreate-ui workflow I described above.
Parcel MCP handles packaging and distribution. Your assistant can create Parcel configurations from your .NET projects, set up code signing and notarisation for macOS, configure Azure Trusted Signing for Windows, and build installers for every platform. Describe what you want in plain English and the AI handles the configuration. No more wrestling with signing certificates at 11pm before a release.
Together, these three servers cover build, debug, and ship. That's not a chatbot bolted onto your IDE. That's an AI-powered workflow across the full development lifecycle.
DevTools MCP and Parcel MCP are part of Avalonia Accelerate and require a licence. Build MCP is completely free, no licence key, no sign-up, no usage limits, because we believe that every developer building with Avalonia deserves an AI assistant that actually knows the framework.
Works Where You Work
We've built setup support for every major AI-enabled editor and CLI tool: VS Code with GitHub Copilot, Visual Studio, JetBrains Rider, Cursor, Windsurf, Claude Code, Claude Desktop, and Gemini CLI all have documented configuration paths. In most cases it's a single JSON snippet or a one-line terminal command.
For VS Code, you can add it through the command palette (search for "MCP: Add Server," choose HTTP, paste the URL) or drop a config file into .vscode/mcp.json. Rider users can go through the AI Assistant settings or create an .idea/mcp.json. Claude Code users get the most satisfying setup: claude mcp add --transport http avalonia-docs https://docs-mcp.avaloniaui.net/mcp. One command. Done. Go build something.
A Note on AI and Frameworks
You might have noticed that we don't lead with "AI-ready!" on our homepage or plaster it across every piece of marketing material. That's deliberate.
AI tooling support is something we consider table stakes. Every serious framework should work well with AI assistants, and we don't typically think of it as a feature worth trumpeting. We've invested heavily in Build MCP, DevTools MCP, and Parcel MCP because they genuinely make developers more productive.
The reason people choose Avalonia hasn't changed. It's the rendering engine. It's genuine cross-platform reach across desktop, mobile, web, and embedded. It's the depth of the control library and the styling system. It's the fact that companies like JetBrains, Unity, and Devolutions trust it for production applications used by millions. AI tooling makes all of that easier to work with, and that's genuinely valuable, but it's one feature in a much more compelling story.
We'd rather be known for building a great framework that happens to have excellent AI tooling than for building AI tooling that happens to sit on top of a framework. The substance comes first. Everything else follows from that.
Get Started
Head over to the Build MCP documentation for step-by-step setup instructions for your editor of choice. You'll be up and running in under a minute, and your AI assistant will finally know what it's talking about when it comes to Avalonia.
If you're already an Accelerate subscriber, set up DevTools MCP and Parcel MCP while you're at it. The three servers together give your assistant the ability to learn the framework, see your application, and ship it. Try the recreate-ui prompt with a screenshot of something you'd like to build. I'm genuinely curious to see what you create.
