Merlin the Wizard — your desktop's been waiting

Clippy grew up.
Merlin learned to think.

The wizard from 1997 is back — wired to modern LLMs, fluent in voice, and aware of what's on your screen. Floats on your desktop. Bring your own brain.

Windows 10 / 11 · ~200MB · MIT licensed

An assistant that remembers where it came from.

Microsoft Agent shipped in 1997 and quietly disappeared in 2009. Clippy got the cultural memory; Merlin got the cultural shrug. We thought he deserved better — so we brought him back, this time with a brain capable of holding a conversation.

Merlin floats on your desktop the way he always did. Click him, talk to him, let him watch what you're working on. Under the cape: any LLM you like — local via Ollama, cloud via Anthropic / Groq / OpenRouter, or self-hosted via Hermes.

Everything the original wished it could do.

Six things Merlin does well. None of them require a corporate subscription, and most of them run on your machine.

🧠

Multi-brain switching

Pick any LLM and swap on the fly. Local models run private and offline; cloud models give you frontier capability.

Anthropic Groq OpenRouter Ollama MiniMax Hermes

🎙

Voice in & out

Whisper STT in, three TTS engines out — gestures sync to whatever Merlin happens to be saying.

ElevenLabs Edge TTS Whisper

👁

Eyes on your screen

Capture any region and ask about it. Pull in web search when local context isn't enough. He sees, he searches.

Screen capture Web search tool

✦

Swappable characters

clippyjs engine Custom sprites

⌨

Tray & hotkeys

Lives in the system tray with submenus for voice, brain, and character. Global hotkeys to summon him on demand.

System tray Global hotkeys

⚙

Extensions

New in v0.4 — extend Merlin with scripted abilities, new tools, and custom brain controllers. The cape has pockets.

Plugin API v0.4+

Local, cloud, or self-hosted.

Merlin doesn't ship with a brain — you bring your own. Here's the quick comparison.

Brain	Type	Cost	Privacy	Best for
Ollama	Local	Free	Total	Private chat, offline use
Hermes Agent	Self-hosted	Free	Total	Power users, custom rigs
Anthropic	Cloud	API key	Sent to provider	Reasoning, long context
Groq	Cloud	Free tier	Sent to provider	Realtime voice replies
OpenRouter	Cloud	Pay-per-use	Sent to provider	Trying many models
MiniMax	Cloud	API key	Sent to provider	Multilingual workloads

Up and casting in 90 seconds.

No CLI required. Pick a brain when the wizard asks, then start talking.

Install

Grab the latest .exe from GitHub Releases and run it. Merlin appears on your desktop, waiting.

Run the Setup Wizard

First-time setup walks you through picking a brain provider, dropping in API keys, and choosing a voice.

Merlin First-Time Setup Wizard welcome screen

Just talk to him

Click Merlin, hold to speak, or use the global hotkey. Ask about your screen, search the web, swap characters.

Merlin floating on the desktop with a yellow speech bubble asking "What can I help you with, traveler?"