Higgs: A Local-First AI CLI for Searching and Classifying Your Proton Mail
Some links in this guide are affiliate links. If you sign up through them, Day Zero Guides may earn a commission at no extra cost to you. This never affects which products we cover or what we say about them. See our affiliate disclosure for details.
What is Higgs?
Higgs is a command-line tool that lets you classify, search, and ask natural-language questions about your Proton Mail inbox — entirely on your own machine. No cloud API calls, no telemetry, no third-party model provider seeing your email content. It relies on self-hosted language models (think Ollama or similar local runtimes under the hood) to do the actual reasoning, while Higgs itself handles the Proton Mail integration, indexing, and query layer.
It's released under Apache-2.0, so it's free to use, fork, and modify. There are no pricing tiers because there's no product to buy — you're running the software yourself, on your own hardware, against your own model weights.
This matters specifically for Proton Mail users, because Proton's whole pitch is privacy — end-to-end encryption, no ads scanning your inbox, Swiss jurisdiction. Piping that mail into a cloud LLM for "smart search" would undercut the reason most people chose Proton in the first place. Higgs is built for people who want AI-assisted email triage without breaking that privacy model.
What it actually does
From the CLI, Higgs appears to handle three core jobs:
- Classification — sorting incoming or existing mail into categories (receipts, newsletters, personal correspondence, action-required, etc.) using a local model's judgment rather than static keyword rules.
- Search — semantic search across your mailbox, so you can find "that email about the apartment lease renewal" without remembering exact subject lines or sender addresses.
- Q&A — asking questions like "what did I agree to pay for the conference sponsorship" and getting an answer synthesized from relevant emails, again computed locally.
Because everything runs on-device, performance depends heavily on your hardware and which model you point Higgs at. A small quantized model on a laptop CPU will be noticeably slower and less accurate than a mid-size model on a machine with a decent GPU. Higgs doesn't ship a model — it's the orchestration layer that sits between Proton Mail and whatever local LLM you've already got running.
Concrete use cases
- Freelancers hunting for old invoices: instead of scrolling through months of Proton Mail, ask "which clients still haven't paid the Q3 invoice" and get a synthesized answer pulled from actual email content.
- Privacy-conscious professionals (lawyers, healthcare, journalists) who use Proton Mail specifically to avoid cloud exposure, and want AI search without sending client correspondence to OpenAI or Anthropic's servers.
- Self-hosters already running Ollama for other local-AI workflows who want to extend that same infrastructure to their inbox instead of adopting a separate SaaS email assistant.
- Digital minimalists doing inbox cleanup who want automatic classification of years of backlog into actionable folders without granting a third-party app OAuth access to their mail.
What it's not
Higgs is not a hosted service — there's no web dashboard, no mobile app, no "sign up and go." You need to be comfortable with a terminal, installing dependencies, and running (or pointing to) a local model. It's also not a Gmail tool — it's built specifically around Proton Mail's API/bridge setup, so if your mail lives elsewhere, this isn't for you today.
How it compares
| Higgs | Proton Mail built-in search | Ollama | Thunderbird + local AI plugins | |
|---|---|---|---|---|
| Price | Free, open source (Apache-2.0) | Included with Proton Mail plan | Free, open source | Free (Thunderbird) + plugin cost varies |
| Core function | AI classify/search/Q&A over Proton Mail via local LLM | Keyword and filter-based search, no AI reasoning | General-purpose local LLM runtime, not email-specific | Email client with optional AI plugins bolted on |
| Privacy model | 100% local inference, no telemetry, no cloud calls | Encrypted by Proton, but no semantic AI layer | Local, but you build the integration yourself | Depends entirely on the plugin — many still call cloud APIs |
| Setup effort | CLI install + point at a local model (e.g. Ollama) | None — already built into Proton Mail | Moderate — install runtime, pull models | Moderate to high — client install plus plugin config |
| Best for | Proton Mail users wanting AI search without leaving the privacy model | Users fine with basic filters, no AI features needed | Developers wanting a general local-LLM backend for anything | Users wanting a full email client with AI features baked into the UI |
Should you try it today?
If you're already a Proton Mail user with a local LLM running via Ollama (or similar), Higgs is a low-risk, zero-cost experiment — it's open source, doesn't phone home, and directly extends infrastructure you may already have. The honest caveat: this is a new, CLI-only tool, so expect rough edges, limited documentation, and dependence on the quality of whatever model you point it at. If you want a polished, GUI-driven AI email assistant, this isn't it yet. If you want to keep AI-assisted email search entirely off the cloud and you don't mind a terminal, Higgs is worth cloning and testing against a week of real inbox traffic before deciding whether it replaces Proton's native search for you.