Skip to main content
AIDiveForge AIDiveForge
Save tools:Log inSign up
Visit Higgs

Get This Tool

License: Apache-2.0 Any use incl. commercial
Local-run terms: Users may install via brew, go install, or release tarballs and run the binary locally with any self-hosted OpenAI-compatible model server.

Share This Tool

Compare This Tool
📋 Embed this tool on your site

Copy this code to embed a compact tool card:

Higgs

FreeOpen SourceSelf-HostedAgentic

Pricing

Model
Free

Summary

Piping Proton Mail to a cloud AI API undoes end-to-end encryption in one request — every subject line, every receipt, every security alert handed to a model provider. Higgs exists to close that gap.

Higgs is an Apache-2.0 CLI that runs AI classification, search, and Q&A against your Proton Mail inbox without a single byte leaving 127.0.0.1. Mail flows through Proton Mail Bridge; inference runs through Ollama or any OpenAI-compatible local server like llama.cpp. Output is NDJSON with a guaranteed terminator and typed exit codes, so agents driving it branch on structured error keys rather than parsing English phrases and hoping. The ceiling appears fast: Higgs is Proton Mail-only, has no API surface, and every workflow requires a locally hosted model — teams without the hardware or patience to run Ollama are stopped before the first classified message.

Bottom line: Pick Higgs when privacy is a hard constraint and you already run Ollama on the same machine as Proton Mail Bridge — skip it when your inbox is on Gmail, or when your team needs a shared inference server rather than per-developer local setup.

Community Performance Report Card

No community ratings yet. Be the first to rate this tool!

Best For: Users who want private AI on Proton Mail, Agent frameworks that need typed CLI tools, Self-hosted setups avoiding any cloud inference, Privacy-focused workflows with structured data output
  • No cloud inference path exists in the codebase, which means Proton Mail's end-to-end encryption is not undermined the moment you add AI to your workflow.
  • NDJSON output with a guaranteed `{"type":"summary"}` terminator, so agent frameworks driving Higgs branch on structured data rather than parsing prose and guessing when a stream ends.
  • Provider-agnostic local inference via any OpenAI-compatible server (Ollama by default, llama.cpp via PM_LLM_BACKEND=openai), so swapping models is a one-line environment variable change rather than a code change.
  • `higgs schema` returns a JSON manifest of every subcommand's flags, stdout format, and typed exit codes, which means agent integrations don't need to re-parse --help or hardcode flag syntax every session.
  • Bridge credentials live in the OS keyring with AES-256-GCM fallback, so sensitive credentials are never passed into a model's context window — the failure mode where secrets appear in prompt logs is closed by design.
  • Proton Mail Bridge is a hard dependency and not an abstraction — Gmail, Fastmail, and any provider other than Proton are unsupported. Teams whose inbox is not on Proton Mail have no migration path within Higgs; they switch to a provider-agnostic tool or build their own IMAP layer.
  • Running inference requires a locally hosted model the user maintains — Ollama or llama.cpp must be installed, running, and loaded with a model before any Higgs command works. Teams without a local machine capable of running a model large enough for email classification are blocked entirely; there is no fallback to a managed endpoint.
  • There is no API surface. Higgs is a CLI, and that is the only interface. Teams that need to embed inbox AI in a web application, a shared service, or a scheduled cloud job cannot expose Higgs as a library or service — at that point they are building a wrapper around a local process, which breaks the self-hosted model anyway.

About

Platforms
macOS, Windows, Linux
API Available
No
Self-Hosted
Yes
Last Updated
2026-08-14T01:51:51.013Z

Best For

Who it's for

  • Users who want private AI on Proton Mail
  • Agent frameworks that need typed CLI tools
  • Self-hosted setups avoiding any cloud inference
  • Privacy-focused workflows with structured data output

What it does well

  • Classify inbox messages with local LLM labels
  • Search and archive mail older than a date
  • Pipe structured output between commands for batch operations
  • Run Q&A and summaries on personal mail locally

Integrations

Proton Mail BridgeOllamallama.cppany OpenAI-compatible server
Help improve this page

Add notes, reviews, and benchmarks so the next visitor gets a clearer picture.

Sign in to contribute

Spotted incorrect or missing data? Join our community of contributors.

Sign Up to Contribute

Frequently Asked Questions

Is Higgs free?
Yes — Higgs is fully free to use. There is no paid tier.
Is Higgs open source?
Yes. Higgs is open source.
Can I self-host Higgs?
Yes. Higgs supports self-hosting on your own infrastructure.
What platforms does Higgs support?
Higgs is available on: macOS, Windows, Linux.

Keeping Proton Mail private when adding AI

Piping Proton Mail to a cloud AI API undoes end-to-end encryption in one request — every subject line, every receipt, every security alert handed to a model provider. Higgs exists to close that gap.

How it works

Higgs is an Apache-2.0 CLI that runs AI classification, search, and Q&A against your Proton Mail inbox without a single byte leaving 127.0.0.1. Mail flows through Proton Mail Bridge; inference runs through Ollama or any OpenAI-compatible local server like llama.cpp. Output is NDJSON with a guaranteed terminator and typed exit codes, so agents driving it branch on structured error keys rather than parsing English phrases and hoping.

Key capabilities

The vendor states use cases include classifying inbox messages with local LLM labels, searching and archiving mail older than a date, piping structured output between commands for batch operations, and running Q&A and summaries on personal mail locally. Platforms supported are macOS, Windows, and Linux. Integrations cover Proton Mail Bridge, Ollama, llama.cpp, and any OpenAI-compatible server. It is free and offers a self-hosted option.

Trade-offs

No cloud inference path exists in the codebase, which means Proton Mail’s end-to-end encryption is not undermined the moment you add AI to your workflow. NDJSON output with a guaranteed terminator lets agent frameworks branch on structured data. Provider-agnostic local inference via any OpenAI-compatible server means swapping models is a one-line environment variable change. On the downside, Proton Mail Bridge is a hard dependency. Gmail, Fastmail, and any provider other than Proton are unsupported. Running inference requires a locally hosted model the user maintains before any command works.

Who it is for / who should skip it

Higgs suits users who want private AI on Proton Mail, agent frameworks that need typed CLI tools, self-hosted setups avoiding any cloud inference, and privacy-focused workflows with structured data output. Teams without Proton Mail or without hardware to run a local model should skip it.