I built Wezzly Companion because I was exhausted by the copy-paste dance with AI or not being able to upload videos and going back and forth 4-5 times before I get to the point!
Every time I wanted help debugging, I’d screenshot my terminal, paste it into ChatGPT, wait, then manually apply the fix. Repeat dozens of times a day.
So I built an AI that just… sees what I’m doing in real time.
How it works,
Runs as a lightweight Tauri desktop app (Rust backend, ~15MB)
Continuously captures your screen (configurable interval)
Sends context to your choice of AI: OpenAI, Claude, Gemini, Grok, DeepSeek, or local via Ollama
Can also execute commands, click, and type — full agent mode if you want it.
All within the Wezzly companion desktop assistant, visible blob that lives on your screen with you and gets stuff done and makes jokes.
How I use it daily
“What’s wrong with this error?” → it reads my terminal
“Summarize this video” → it watches with me
“Help me fill out this form” → it sees the webpage
No screenshots. No explanations. No copy-paste.
Privacy
Everything runs locally.
Your API keys stay on your machine (system keychain).
Your screen data never touches my servers.
Tech - Tauri + Rust + TypeScript
macOS signed & notarized
Windows build available (unsigned for now)
Open-source pledge
I plan to open-source at 10k users.
Keeping it closed for now to protect early development.
Hey HN!
I built Wezzly Companion because I was exhausted by the copy-paste dance with AI or not being able to upload videos and going back and forth 4-5 times before I get to the point!
Every time I wanted help debugging, I’d screenshot my terminal, paste it into ChatGPT, wait, then manually apply the fix. Repeat dozens of times a day.
So I built an AI that just… sees what I’m doing in real time.
How it works, Runs as a lightweight Tauri desktop app (Rust backend, ~15MB)
Continuously captures your screen (configurable interval)
Sends context to your choice of AI: OpenAI, Claude, Gemini, Grok, DeepSeek, or local via Ollama
Can also execute commands, click, and type — full agent mode if you want it.
All within the Wezzly companion desktop assistant, visible blob that lives on your screen with you and gets stuff done and makes jokes.
How I use it daily “What’s wrong with this error?” → it reads my terminal “Summarize this video” → it watches with me “Help me fill out this form” → it sees the webpage
No screenshots. No explanations. No copy-paste.
Privacy Everything runs locally. Your API keys stay on your machine (system keychain).
Your screen data never touches my servers.
Tech - Tauri + Rust + TypeScript
macOS signed & notarized
Windows build available (unsigned for now)
Open-source pledge I plan to open-source at 10k users. Keeping it closed for now to protect early development.
Demo: https://youtu.be/ya1Kz_iAraE
Would love feedback — what would make this useful in your workflow?