Sender view

Other email clients reason over messages. mxr also reasons over people. The sender view answers questions like:

Who emails me most?
Who do I reply to fastest?
Which senders have I left hanging?
Did this person ask me anything I haven’t responded to?

These questions are cheap when your archive is on disk and indexed. They’re impossible when your archive is behind a third-party API and a per-call quota.

CLI

mxr sender alice@example.com

Prints a profile:

Volume — messages over time, inbound vs outbound
Cadence — median response latency in both directions, both clock-time and business-hours
Open threads — threads where the most recent message is theirs and you haven’t replied
Open commitments — questions you’ve asked them, sorted by age
Recent activity — the last N messages from that sender, including thread ids for direct navigation in clients
Relationship memory — local style, summaries, known topics, and open commitments when the background profile has been built

mxr sender alice@example.com --format json | jq .

Web app

In the web app, click Sender while reading a message. The right rail shows the sender profile as structured sections:

Identity and relationship stats
Inbound/outbound volume and reply cadence
Storage and attachment-heavy context when available
Other emails from sender — recent messages from the same sender, excluding the currently open thread

Each item in Other emails from sender links directly to that conversation, so you can jump to the mail you remember without building a manual search query.

Common workflows

Find people I owe replies to

mxr stale --mine --older-than-days 7 --format json \
  | jq -r '.threads[] | "\(.days_stale)\t\(.subject)"' \
  | head

Then run mxr sender <email> on each to confirm before triaging.

Pick the “biggest” senders interactively

mxr storage --by sender --format jsonl \
  | jq -r '"\(.size_bytes)\t\(.label // .key)"' \
  | sort -rn \
  | fzf --header='bytes | sender' \
  | awk '{print $2}' \
  | xargs mxr sender

storage ranks senders by data weight; sender opens their profile.

Daily standup: who emailed me overnight?

mxr search 'newer_than:1d' --format json \
  | jq -r 'group_by(.from)
           | map({sender: .[0].from, count: length, latest: max_by(.date).subject})
           | sort_by(-.count) | .[]
           | "\(.count)\t\(.sender)\t\(.latest)"'

TUI

Inside the TUI, open the sender profile modal for the focused message:

Ctrl-p (palette) → “Sender View”
The modal shows the same profile sections plus Other emails from sender.
Use j / k to move through that list, then Enter or o to open the selected conversation.
“Open in CLI” copies the equivalent mxr sender ... invocation to the clipboard.

Why this matters for agents

A sender profile is the most useful single context an LLM can have when drafting a reply or evaluating a triage candidate:

mxr sender alice@example.com --format json \
  | claude -p "Based on cadence and recent threads, draft a friendly nudge if she's overdue."

The agent gets relationship shape as JSON, without needing to read every email.

Relationship-aware drafting

Draft assist can use sender profile data as weak background context. It helps match cadence and tone, but it is not treated as truth over the current thread. The prompt explicitly tells the model that current message content and your instruction win, and that anything not listed as a known topic is unknown.

Generated drafts return humanizer and voice-match metadata so clients can warn when a reply sounds robotic or drifts away from the known relationship style.