Tool Lab

LOCAL FILE UTILITY · U3

MBOX label search and completeness summary

Scan mail archives in bounded chunks, filter labels and headers, and export summaries or matching original messages.

Scope: Common ctime-style mboxo/mboxrd; first 512 MiB. Raw UTF-8 header search; not forensic proof.

Search an MBOX archive locally

Choose an mboxo/mboxrd archive with common ctime-style “From ” separators. Scanning reads 1 MiB chunks and keeps bounded summaries. Messages and bodies stay in this browser; no account connection is needed.

Optional filters are case-insensitive substrings, combined with AND. Search uses raw UTF-8 headers; MIME encoded-word headers are not decoded. Label counts cover all scanned messages.

No archive scanned yet.

Limits: first 512 MiB, 1,000,000 finalized messages, 10,000 retained matching ranges, 5,000 distinct labels, 100 labels per message, 1,024 characters per counted label, 64 KiB per line and 256 KiB per message header. Split larger archives or narrow filters when a limit is reported. Unfinished trailing messages are excluded from matching export; malformed matches and truncated match lists block that export. A missing final newline is flagged as possible truncation. Structural checks cannot prove forensic completeness or distinguish an unescaped body line identical to a valid separator. Unsupported MBOX dialects require re-export.
Summary downloads include labels and structural counts, with sender/subject values and message bodies omitted. Matching MBOX downloads contain the selected original messages, including their private contents. Review partial-scan warnings before using any export.
How to use
  1. Choose MBOX and optional label, From and subject filters.
  2. Scan locally and review byte counts, malformed records and limits.
  3. Download summaries or retained matching messages.
Limits and context
Read format limits and warnings before exporting. Sample buttons use synthetic content. No uploaded content is sent to a server.
Original task discussion · Implementation context
The discussion explains task context; it does not prove this tool is better than existing alternatives.