Skip to content

mxr assistant

Generated from mxr assistant --help. After editing crates/daemon/src/cli/, update and review the snapshots with scripts/cargo-test -p mxr --test cli_help, then rebuild site/.

Talk to the assistant about your mail. It reads and searches freely; every change it wants to make waits for your approval

mxr assistant [OPTIONS] <COMMAND>
Subcommand Purpose
status Whether the assistant can run, and on which model
model The model the assistant uses: list the choices, or set one for every conversation. Hosted models need their key (mxr llm key set)
send Send a message, starting a conversation unless –conversation is given, and wait for the reply. Changes the assistant wants to make, and questions it asks, are printed with the command that answers them
approve Approve a change the assistant is waiting on. It runs exactly the call shown
reject Decline a change or question; nothing runs
answer Answer a question the assistant asked. For one question, use –choose and/or –other; for several, –answers-json
trust Let reversible changes (archive, move, star, read, labels) run without asking in one conversation. Sending, deleting, trash, spam and rules always ask
stop Stop a running turn, or close what is waiting without running it
list Conversations, most recently active first
show Show a whole conversation
delete Delete a conversation and its transcript
help Print this message or the help of the given subcommand(s)
Flag Description
--format <FORMAT> [possible values: table, json, jsonl, csv, ids]
-h, --help Print help

Check assistant availability and inspect conversations before continuing one. Deleting a transcript supports a preview.

Terminal window
mxr assistant status --format json
mxr assistant list --limit 10 --format json
mxr assistant show CONVERSATION_ID --format json
mxr assistant delete CONVERSATION_ID --dry-run

Answer a question the assistant asked. For one question, use –choose and/or –other; for several, –answers-json

mxr assistant answer [OPTIONS] <CONVERSATION> <CALL_ID>
Flag Description
<CONVERSATION>
<CALL_ID>
Flag Description
--choose <CHOOSE> An option’s label (repeatable for multi-select questions)
--other <OTHER> Your own answer
--answers-json <ANSWERS_JSON> One answer per question: [{“selected”: [“Work”], “other”: null}, …]
--no-wait
--format <FORMAT> [possible values: table, json, jsonl, csv, ids]
-h, --help Print help

Approve a change the assistant is waiting on. It runs exactly the call shown

mxr assistant approve [OPTIONS] <CONVERSATION> <CALL_ID>
Flag Description
<CONVERSATION>
<CALL_ID>
Flag Description
--no-wait
--format <FORMAT> [possible values: table, json, jsonl, csv, ids]
-h, --help Print help

Delete a conversation and its transcript

mxr assistant delete [OPTIONS] <CONVERSATION>
Flag Description
<CONVERSATION>
Flag Description
--dry-run Show what would be deleted, without deleting it
--format <FORMAT> [possible values: table, json, jsonl, csv, ids]
-h, --help Print help

Conversations, most recently active first

mxr assistant list [OPTIONS]
Flag Description
--limit <LIMIT>
--format <FORMAT> [possible values: table, json, jsonl, csv, ids]
-h, --help Print help

The model the assistant uses: list the choices, or set one for every conversation. Hosted models need their key (mxr llm key set)

mxr assistant model [OPTIONS] [COMMAND]
Subcommand Purpose
set Use this model for the assistant’s next message, in every conversation
help Print this message or the help of the given subcommand(s)
Flag Description
--format <FORMAT> [possible values: table, json, jsonl, csv, ids]
-h, --help Print help

Use this model for the assistant’s next message, in every conversation

mxr assistant model set [OPTIONS] <MODEL>
Flag Description
<MODEL> A model id from mxr assistant model, such as claude-haiku-5-5
Flag Description
--dry-run Check the choice and show what would change, without changing it
--format <FORMAT> [possible values: table, json, jsonl, csv, ids]
-h, --help Print help

Decline a change or question; nothing runs

mxr assistant reject [OPTIONS] <CONVERSATION> <CALL_ID>
Flag Description
<CONVERSATION>
<CALL_ID>
Flag Description
--reason <REASON> Tell the assistant why, or what to do instead
--no-wait
--format <FORMAT> [possible values: table, json, jsonl, csv, ids]
-h, --help Print help

Send a message, starting a conversation unless –conversation is given, and wait for the reply. Changes the assistant wants to make, and questions it asks, are printed with the command that answers them

mxr assistant send [OPTIONS] <MESSAGE>
Flag Description
<MESSAGE>
Flag Description
--conversation <CONVERSATION> Continue this conversation
--thread <THREAD> The thread you mean by “this thread”
--message <MESSAGES> Messages you mean by “these” (repeatable)
--no-wait Return as soon as the turn starts
--format <FORMAT> [possible values: table, json, jsonl, csv, ids]
-h, --help Print help

Show a whole conversation

mxr assistant show [OPTIONS] <CONVERSATION>
Flag Description
<CONVERSATION>
Flag Description
--format <FORMAT> [possible values: table, json, jsonl, csv, ids]
-h, --help Print help

Whether the assistant can run, and on which model

mxr assistant status [OPTIONS]
Flag Description
--format <FORMAT> [possible values: table, json, jsonl, csv, ids]
-h, --help Print help

Stop a running turn, or close what is waiting without running it

mxr assistant stop [OPTIONS] <CONVERSATION>
Flag Description
<CONVERSATION>
Flag Description
--format <FORMAT> [possible values: table, json, jsonl, csv, ids]
-h, --help Print help

Let reversible changes (archive, move, star, read, labels) run without asking in one conversation. Sending, deleting, trash, spam and rules always ask

mxr assistant trust [OPTIONS] <CONVERSATION>
Flag Description
<CONVERSATION>
Flag Description
--off Ask before every change again
--format <FORMAT> [possible values: table, json, jsonl, csv, ids]
-h, --help Print help