Summaries are only useful if you can check them. An assistant that tells you "the vendor agreed to the revised deadline" without pointing at the message where that happened has given you a claim, not an answer. This post explains how mailin's AI Assistant and Thread Story are built around citations, what the AI Digest adds, and what the free and Personal tiers include.
Why an AI email assistant should summarize threads with citations
An email archive is a record. When you ask a question about it, the answer matters less than the evidence behind it: which message, from whom, on what date. A lawyer needs to quote the email, not the summary. An investigator needs the Message-ID. A journalist needs to know that a quote came from the archive and not from the model's idea of what a plausible email says.
mailin's approach is that every answer the assistant gives is grounded in specific emails, and those emails are shown alongside the answer. You can open each one, read it in full, and decide for yourself whether the summary holds up.
The AI Assistant: plain-language questions, cited answers
You ask in ordinary language: who first raised the pricing change, what was decided about the office move, which threads with a particular domain mention a contract. The assistant searches your archive, drafts an answer, and cites the exact emails it drew on.
The answer is generated by Apple Intelligence, Apple's on-device Foundation Models, so the question and the archive never leave the device. That engine needs macOS 26 or later, or iOS 26 or later, on a supported device. On older systems, mailin falls back to Apple NaturalLanguage, which covers sentiment, language detection and named-entity recognition but does not generate written answers. The on-device AI post goes into the two engines in detail.
Thread Story: a conversation as a cited timeline
Long threads are hard to read in a mail client. Replies fork, subject lines drift, people are added and dropped, and the quoted text at the bottom of each message repeats what came before. Thread Story rebuilds the whole conversation with one click as a chronological timeline, each entry tied to the email it came from.
The threading itself is deterministic: mailin follows the Message-ID, In-Reply-To and References headers to establish which message replied to which. That part does not involve AI at all, so the order of events is not something a model guessed.
On top of the timeline you can ask for an optional narrative. It is generated on-device and grounded only to the emails in that thread. It cannot draw on other threads, the wider archive or anything outside the conversation, which keeps it from blending unrelated messages into one story. Because the narrative is grounded to the same emails that make up the timeline, you can check it line by line against the entries below it.
Digests: what happened, and what your saved searches turned up
Two related features work on a time range rather than a single question. The daily or weekly AI Digest summarizes the important emails from a period: sentiment trends, urgent items and the key threads. It is a way to get oriented in an archive before you start searching.
Separately, any search you build with mailin's operators and filters can be saved, and you can opt into a weekly digest of your saved searches. If you are monitoring an archive for a set of terms, senders or patterns, that digest is where new matches surface.
Verify before you rely on it
AI features can produce inaccurate or incomplete results. The citations exist so you can check every claim against the source; use them, especially before quoting a summary in a report, a filing or an article. If an answer cites three emails, read the three emails.
Two more details help with that. Every AI output carries an AIProvenance tag naming the engine that produced it, so you know whether a summary came from Apple Intelligence, NaturalLanguage or the optional Cloud AI mode. And Cloud AI is off by default: if you enable it with your own OpenAI or Anthropic key, only the excerpts you choose are sent to your provider, never to the developer, and each feature asks for consent separately.
What each tier includes
The free tier covers up to 500 emails and includes three AI assistant queries and three AI smart-filter queries, plus on-device sentiment and topic analysis. That is enough to import a sample and see how a cited answer looks against your own mail.
The Personal tier removes the caps: unlimited emails, unlimited AI, the Apple Intelligence AI Assistant, daily AI digest and the smart tagger. It is $4.99 per month, $44.99 per year or $99.99 as a one-time lifetime purchase. The Professional tier adds the forensic and legal tooling, including predictive coding that learns relevance as you review. Those features are designed to support common records-integrity and eDiscovery workflows; admissibility of digital evidence is jurisdiction-specific and depends on factors beyond any single software tool, so consult qualified legal counsel for evidentiary use. Full details are in the pricing section.
FAQ
Does the assistant need an internet connection?
No. The default engines run on the device, and mailin works fully offline. A connection is used only if you deliberately enable the optional Cloud AI mode with your own API key.
Can Thread Story handle a thread where the subject line changed?
Threading follows the Message-ID, In-Reply-To and References headers rather than the subject line, so replies stay linked even when someone edits the subject.
What if I am on macOS 14.6 or iOS 17.6?
Sentiment, language detection and entity extraction work on every supported device through Apple NaturalLanguage. The written summaries, meaning the assistant's answers, the Thread Story narrative and the AI Digest, depend on Apple Intelligence, which requires macOS 26 or later or iOS 26 or later.