Feeding mailbox archives (.mbox) to an LLM
An .mbox is a whole mailbox in one file — every
message concatenated, each with its own MIME plumbing and quoted
history. It's what "Export" gives you in many mail tools and what
Google Takeout produces. Ideal source material for "what did we
decide?" questions, unreadable as-is.
The element mapping
| In the .mbox | In the Markdown |
|---|---|
| Each message | ## Message N · {subject} — an addressable thread outline |
| Headers / encodings / quotes | Exactly the .eml treatment, per message |
| Archives beyond 50 messages | First 50 converted, the cut stated explicitly |
| Attachments across the archive | Counted and listed per message, not extracted |
Before → after
In the file
From jordan@… Thu Jul 2 09:14:22 2026
Subject: Kickoff
…
From sam@… Thu Jul 2 17:02:10 2026
Subject: Re: Kickoff
> (the whole thread again…)In the Markdown
## Message 1 · Kickoff
- **From:** Jordan Lee …
## Message 2 · Re: Kickoff
Shortlist should be two vendors.
> … (9 more quoted lines truncated)FAQ
How do I get an .mbox? Google Takeout (Gmail), Thunderbird folder export, Apple Mail "Export Mailbox".
Huge archives? Fifty messages per conversion — split the file for more (the output tells you when it cut).
Private mail? Local conversion, nothing uploaded — verifiably.
Export a thread and see it become an addressable outline.