I've been working on SpamBuster, a free and open-source email spam analyzer that uses AI to intelligently clean your inbox based on custom rules you define.
What it does:
You can connect your email accounts to the software (Gmail via Google API, Outlook via Microsoft API or any IMAP server) and connect to Ollama or OpenRouter (more providers coming soon) to analyze your incoming emails according to custom rules. Unlike traditional spam filters, you can configure your rules with natural language that will be sent to the AI along with the email.
Why I built it:
I receive around 5-6 SPAM a day and I'm exhausted about it. I tried Outlook, Gmail, other providers self-hosted emails but all the anti-spams are so useless. Lack of configuration, lack of intelligence, and I still have my inbox full of spam. I looked for AI email software, but they're all about being a complete software with a dashboard managing your emails and more like an AI assistant for your emails. I don't want another software, I want to use my traditional email client but without spam.
Key features:
- Works with Gmail (Google API, compatible with Google Workspace), Outlook (Microsoft API, compatible with Microsoft 365), and any IMAP server
- Uses OpenRouter or Ollama
- Customizable rules - define your own spam detection criteria
- Cost tracking: if you use OpenRouter, the cost associated with each analysis is displayed next to each email, plus there's a stats tab where you can see the cost over time.
- Includes LanceDB (vector database) to match against previously analyzed emails for improved accuracy
- Electron built: Cross-platform desktop app (Linux, macOS, Windows)
- Modern UI built with Next.js and Shadcn UI
Privacy or performance focused:
Choose between privacy and run it fully locally using Ollama, or choose the performance and use OpenRouter.
The best models to use:
I use ChatGPT 5 nano, that is very cheap and efficient. You may use smarter models but be careful with the cost.
If you want to use Ollama, I tried Phi-4 mini. It's great but you may have to be careful on how to write your rules., as the model is less smart than others like chatgpt 5 nano. I didn't try other models, but I guess Gemma or so is good too (I'm looking for feedback on this point).
A word about cost:
ChatGPT 5 nano costs me around a cent a week, and I receive around 10 emails a day in my inbox. Of course it's free if you use Ollama.
License: MIT
Download: Pre-built binaries are available for Linux (AppImage) and SnapCraft, macOS (DMG), and Windows (exe installer)
GitHub: https://github.com/lelenaic/spambuster
I'm mostly looking for feedback at this moment. If you're interested in AI-powered email management or just want to help improve the project, feel free to check it out!
Discussion