Klorn

Most email AI tools automate the sending. Klorn is built around the opposite bet: that the actual bottleneck is figuring out what deserves a decision, not executing after you’ve decided. The core workflow is a four-step loop — observe signals from mail and calendar, condense them into short cards with attached evidence, draft replies or schedule changes as staged drafts, then surface one explicit approval step before anything external happens. The vendor describes this as a ‘decision queue’ rather than an inbox, and a daily log records every action taken.

The differentiating design choice is the approval gate. Each decision card groups related signals — an unanswered thread, a calendar block, and an open task — into a single view with context about risk and next action, then holds that action until you tap Approve, Edit first, or Hold. That means every outbound reply or calendar change is traceable to a specific approval event, which the vendor positions as an auditable log.

Klorn is built for professionals running high-volume correspondence who want AI filtering but not autonomous sending — the docs describe it as explicitly not acting on your behalf. The self-hostable, AGPLv3 open-source release means teams with data-handling requirements can run it on their own infrastructure without touching the paid tier. Where it breaks: teams that need shared queues across multiple users, or workflows where the agent should act without a human in the loop, will find the approval-first architecture a ceiling rather than a feature.

As of the page scraped, Klorn is in invite-only beta with a cap of 100 testers pending a Google CASA security review. Google account integration is required; no API is exposed for external system connections.