Question 1

What can I do with Avalonia and AI?

Accepted Answer

Avalonia ships three Model Context Protocol (MCP) servers that turn any MCP-aware AI assistant into a teammate that understands the framework. Build MCP gives the AI live access to the Avalonia documentation and expert rules so it stops hallucinating APIs. DevTools MCP lets the AI inspect a running Avalonia app, read and change properties, and capture screenshots. Parcel MCP exposes cross-platform packaging, code signing and notarisation as tool calls. Together they cover the full loop: write code, see the running app, ship installers — all driven from a chat with your assistant.

Question 2

Which AI assistants and editors work with Avalonia?

Accepted Answer

All three Avalonia MCP servers work with any client that speaks Model Context Protocol. That includes VS Code with GitHub Copilot, Visual Studio 2022 (17.x and later), JetBrains Rider, Cursor, Windsurf, Claude Code, Claude Desktop and Gemini CLI. You configure each server once in your editor or assistant of choice and the same tools are available everywhere.

Question 3

Why use MCP servers instead of just chatting with an AI about Avalonia?

Accepted Answer

General-purpose models were trained on a snapshot of the internet that's months or years out of date, and they have no way to look at your running app or your build pipeline. The Avalonia MCP servers replace stale training data with live docs, give the AI eyes on the actual visual tree of your app and let it run real builds. The result is fewer hallucinated APIs, suggestions grounded in what your UI is doing right now, and packaging steps that actually execute instead of being described.

Question 4

Do I need an Avalonia subscription to use AI tools with Avalonia?

Accepted Answer

No. Build MCP is free, requires no licence key, and is the recommended starting point. It runs as a hosted remote server, so there is nothing to install — you just point your AI client at it. DevTools MCP and Parcel MCP require an Avalonia Plus subscription or higher because they expose paid product features (live diagnostics and cross-platform packaging respectively).

Question 5

Is using AI with Avalonia secure?

Accepted Answer

The MCP servers run locally inside your editor session (or, in the case of Build MCP, as a remote read-only docs server), so your source code is never sent to Avalonia. DevTools MCP only sees the app instance you explicitly attach it to, and Parcel MCP only touches the project directories you point it at. License keys for the paid servers are read from environment variables on your machine — they never leave it.

Question 6

What is the Avalonia Build MCP server?

Accepted Answer

Build MCP is a free, hosted Model Context Protocol server that gives your AI assistant live access to the Avalonia knowledge base. It exposes five tools: search_avalonia_docs for full-text documentation search, lookup_avalonia_api for API reference lookups, get_avalonia_expert_rules for AXAML, MVVM, binding and styling guidance, migrate_diagnostics for analyser-driven fixes, and migrate_to_xpf for guided WPF-to-XPF migration.

Question 7

How do I install the Build MCP server?

Accepted Answer

There's nothing to install. Build MCP is a remote HTTP server hosted at https://docs-mcp.avaloniaui.net/mcp. You add a single MCP server entry in your editor or assistant pointing to that URL and the tools become available immediately. No NuGet package, no licence key, no local process.

Question 8

Does Build MCP fix the hallucinated API problem?

Accepted Answer

Yes — that is exactly the problem it was built to solve. When the AI needs an Avalonia API, control name or property, it calls lookup_avalonia_api or search_avalonia_docs and gets the answer from the current docs rather than guessing from memory. The expert-rules tool also feeds the assistant Avalonia's authoritative AXAML, MVVM and binding conventions so generated code follows the same patterns your team uses.

Question 9

Can Build MCP help me migrate WPF code to Avalonia or XPF?

Accepted Answer

Yes. The migrate_to_xpf tool walks the AI through a guided WPF-to-XPF migration, and migrate_diagnostics lets the assistant fix code based on Avalonia analyser output. Build MCP also ships pre-written prompts — init, new, recreate-ui and wpf-migration — that your AI client can invoke as starting points.

Question 10

What does the DevTools MCP server let the AI do?

Accepted Answer

DevTools MCP gives an AI assistant live introspection of a running Avalonia app. The AI can walk the visual tree, search by selector, read and modify property values, change pseudo-classes, inspect resolved styles and resources, dispatch input events, and capture a screenshot of any control. It can also attach to a XAML file via the previewer without running the full app, which makes design iteration much faster.

Question 11

What subscription do I need for DevTools MCP?

Accepted Answer

DevTools MCP is part of the Avalonia developer tools, so it requires an Avalonia Plus subscription or higher. The Community edition of DevTools does not include it. The licence key is read from the AVALONIA_TOOLS_LICENSE_KEY environment variable on Avalonia 12.0 and later, or ACCELERATE_LICENSE_KEY on the 11.x line.

Question 12

How do I add DevTools MCP to my Avalonia app?

Accepted Answer

Add the AvaloniaUI.DiagnosticsSupport NuGet package and call .WithDeveloperTools() on your AppBuilder, or .AttachDeveloperTools() on a window during development. Run your app, then start the MCP server from your editor and use the attach-to-app tool to connect it to the running process. From that point on the AI can drive the app.

Question 13

Can the AI actually change my running UI, or just look at it?

Accepted Answer

Both. Read-only tools like tree, ancestors, search, props, styles, screenshot and resources let the AI understand the current state. Active tools — set-prop, pseudo-class, action, input and open-asset — let it change properties, trigger states like :pointerover, dispatch input events and open referenced assets. So the AI can ask 'what does this button look like?', try a fix, and screenshot the result without you ever leaving the chat.

Question 14

What is the Parcel MCP server?

Accepted Answer

Parcel MCP exposes Avalonia Parcel — Avalonia's cross-platform packaging, code-signing and notarisation tool — as MCP tool calls. The AI can create a Parcel configuration from your existing .NET project, set up Windows code signing via Azure Trusted Signing, set up macOS signing with a P12 certificate and notarisation via your Apple ID, and run builds that produce DMG, DEB, NSIS and ZIP installers from a single chat command.

Question 15

Do I need an Avalonia subscription for Parcel MCP?

Accepted Answer

Yes. Parcel and the Parcel MCP server require an Avalonia Plus subscription or higher. The licence key is read from the PARCEL_LICENSE_KEY environment variable. The MCP server is included with the Parcel CLI itself, so once your subscription is active you can use it on every machine you sign in on.

Question 16

How do I install and start Parcel MCP?

Accepted Answer

Install the Parcel CLI as a global .NET tool with `dotnet tool install -g parcel`, then run `parcel mcp` to start the server. Add an MCP server entry in your editor pointing to the local Parcel process and the AI can call setup-trusted-signing, setup-apple-sign, setup-apple-notary, create-project and pack on your behalf.

Question 17

Can the AI ship a signed installer without me leaving my editor?

Accepted Answer

In practice, yes. Once Parcel is configured for your project, the AI can run `pack` for any supported target — Windows, macOS or Linux — and produce a signed, notarised installer. You stay in the chat: ask for 'a signed macOS DMG' or 'a Windows installer with Trusted Signing', the AI calls Parcel, and you get an artefact you can hand to a customer.

Question 18

Where do I start if I just want to try Avalonia with AI?

Accepted Answer

Start with Build MCP. It is free, hosted at https://docs-mcp.avaloniaui.net/mcp, and the only step is registering it as an MCP server in your AI client. From that point your assistant has accurate Avalonia docs, API reference and expert rules — that alone fixes most of the hallucinated-API problem when generating Avalonia code.

Question 19

Can I use all three MCP servers at the same time?

Accepted Answer

Yes, and in practice you usually want to. Build MCP grounds the AI in correct Avalonia knowledge, DevTools MCP lets it see and drive the running app, and Parcel MCP lets it ship the result. They expose disjoint sets of tools so there is no overlap, and most clients will route each request to the right server automatically.

Question 20

Where can I find the official setup instructions for each MCP server?

Accepted Answer

The Avalonia documentation has a dedicated AI tools section at docs.avaloniaui.net/tools/ai-tools/, with one page per server: build-mcp covers the free docs server, developer-tools/mcp covers DevTools MCP, and parcel/mcp covers the Parcel MCP server. Each page has copy-pasteable configuration snippets for VS Code, Visual Studio, Rider, Cursor, Windsurf, Claude Code, Claude Desktop and Gemini CLI.

Develop Avalonia apps with AI.

One protocol, three lenses on your app.

Build MCP

DevTools MCP

Parcel MCP

Two lines, then ask anything.

What we've shipped, what we've learned.

Accelerate Showcase: a desktop app built with AI

Introducing the Avalonia DevTools MCP server

Avalonia and AI,
answered.

Pair with an AI that knows Avalonia.