Ask Mo about your project. Get the answer and the notes it came from.

Mo Assistant is the agent inside Morion. Your coding agents work the code; Mo works the notebook. Ask what was decided and get the note it came from, with a status on every claim. Find that ticket. Write a note from a conversation, set up a folder, tidy the board. Mo is a panel in the app and the same set of tools over MCP, so your agents can ask it too.

cited answersstatusesapproval for risky actions

Ask Mo

Ask Mo · Reminders decisionOpened by you

Ask Mo

AllAsked by meFrom Mo

New chat

Today

Reminders decision

Topic cleanup: onboarding

Yesterday

What did Codex change

Did we decide how reminders work?

Used 3 tools mo_search, notes_get, notes_get

Yes, mostly. Reminders are scheduled locally and mirrored to the server queue on sync Reminder architecture decided. Push delivery is out of scope until v2 Reminder architecture decided.

Whether queued deletes replay in order is still open Queue writes while offline · comments provisional.

Conflict: the March spec says reminders are server-first Sync protocol v1 superseded; the decision note overrides it.

Gap: nothing in the notes covers reminder time zones. Want me to open a ticket for it?

Ask Mo or give him an instruction…

Every claim names the note it came from and carries a status: decided, provisional, proposed, superseded, not implemented, unknown. Contradictions come back as conflicts, missing facts as gaps.

Pick a question. Read the answer with its receipts.

Three questions over a small demo project. Every answer is shaped like a real Mo answer: cited claims, a status per claim, conflicts and gaps called out.

Ask Mo · Habit Tracker4 sources · verified

Reminders are mentioned, not decided.

Check-in flow lists "undo, notes, reminders" as scopedecided
Plan: habit tracker proposes local notifications first, "push later if retention needs it"provisional
No note covers notification permissions on iOSunknown

Gap: nothing decides local vs push, and nothing schedules the permission prompt. Conflicts: none.

How it works

Ask in plain words

Scope to a project, a folder or a set of kanban statuses. "What is left on Offline mode?" "Did we decide how reminders work?" "What did Codex change this week?"

Mo retrieves, then answers

A bounded retrieval loop over the index, then an answer with a citation on every claim. Quotes are checked against the source; a claim that cannot be verified is marked, never silently kept.

Statuses, conflicts, gaps

decided, provisional, proposed, superseded, not implemented, unknown. Two notes that disagree come back as a conflict; a question the notes cannot answer comes back as a gap.

Let Mo do things

Create notes and folders, move cards, edit workflows. Anything destructive waits for your approval in the panel.

In detail

Find

"Find that ticket about reminders" in plain words, or raw hybrid search when an agent wants a ranked list and nothing else.

Memory

Workspace-wide facts you and your agents can add and remove. A conflicting fact is surfaced, never silently overwritten.

Escalation

An agent that needs you leaves a durable question on the ticket or in the backlog. It waits there until you answer. Chats Mo opens for you are marked "From Mo".

Sessions and quick actions

Chats keep their thread. When Mo needs a decision it offers options you can pick or overrule with your own instruction.

Runs on what you already have

Indexing, retrieval and Ask Mo over MCP run on your Claude Code login, or on your Codex login on macOS. The chat panel is part of Pro and runs on OpenRouter, with any model, or on a local Ollama for a fully offline loop.

What it is not

Not a chatbot over your files

Mo answers from the index it built and cites the note. If the notes do not say, Mo says so.

Six jobs, one index.

Each is a tool your agents call over MCP and a thing you can do from the app. Plain sentence first, the tool's real name after it.

Answers with citations and statuses mo_ask

"What did we decide about X?" comes back with the note behind each claim, a status per claim, conflicts between notes, and gaps where the notebook is silent. Scoped to a project, a folder, or the open columns of a board.

Work packets for your agents mo_take_task

One call returns everything around the ticket, and can claim it too: the body, the comments that matter, the project brief, related lessons, specs and rules, and the topics it touches. The agent starts with the story.

Find that ticket, that note, that decision mo_find · mo_search

Conversational lookup that combines keyword and semantic search and says why each hit matched. "Similar to this note" works too. Raw search when an agent only wants a ranked list.

Workspace memory you can read and edit mo_remember · mo_forget

Agents can store a fact for the whole workspace. Mo flags a conflict with what it already knows instead of overwriting silently. You can forget anything, by pattern or all at once.

Asks before anything risky mo_request_human · mo_check_workflow

A question for you lands as a comment on the ticket or a card in the backlog, not in a chat you will scroll past. Mo checks an intended action against your folder rules and says allow, deny, or ask.

Decides between stages in Auto-code mo_stage

Inside an Auto-code workflow every mo_stage node is Mo reading the ticket, the diff and the review, then choosing accept, reject, review, or ask the human.