One notebook and one board for you and every coding agent you use.
Notes, a kanban board, a local MCP server, permissions, smart indexing, Mo Assistant and Auto-code, in one file on your disk. Claude Code, Codex and Cursor read and write the same notes and tickets. Each section shows the screen as it looks in the app.
Notes in Markdown that you and your agents edit on the same page.
Folders, tags and Markdown bodies in one file on your disk. You write in the app; Claude Code, Codex or Cursor create, append to and comment on the same notes over MCP. The activity feed shows who wrote what, in order: you, Claude Code, Codex, Mo.
- Nested folders, tags, pinned notes, trash and archive
- Revisions named by author; comment threads with an AI marker on agent replies
- Full-text and semantic search, in the app and over MCP
- Import from Markdown, text, Word and Apple Notes; export to .md in one click, never gated
Reminder architecture
Reminders are scheduled locally and mirrored to the server queue on sync. Decided: local first, server as backup. Push delivery stays out of scope until v2.
Open question: do queued deletes replay in order? See Queue writes while offline.
A board where agents claim tickets and you approve at Review.
Any note can become a task. Say “take the next task” and the agent claims one ticket, moves it to Doing and reports progress as comments on the card. It stops at Review; you approve, or send it back with a reason.
- Statuses agents understand: backlog · todo · doing · review · Re-opened · done
- Atomic claiming: two agents never take the same card
- Re-opened carries the reason and the next step: resume the session, ask a human, or start fresh
- A private board for a solo founder; in a team it sits below Jira, Linear or ClickUp
Every agent on your machine works the same notes and tasks, in parallel.
Morion runs a local MCP server. Claude Code, Codex, Cursor, Cline, Zed and any other MCP client connect to it at the same time and get the same tools over the same data. Nothing leaves your computer: no server to host, no key to paste.
- Tools for notes, folders, tags, tasks, workflows and Mo, from any agent session
- Several sessions on one board at once; every write logged with the actor
- Any client that speaks MCP, Gemini Antigravity included
- The full tool reference is on the MCP page
| Actor | Calls | Last seen |
|---|---|---|
| mcp:claude-code | 1,204 | 2m ago |
| mcp:codex | 318 | 41m ago |
| mcp:cursor | 57 | Yesterday |
Connect Claude Code, Codex, Cursor and the other agents on your machine in one click.
Morion finds the coding agents installed on your machine and writes the MCP entry, backing up the config file first. The Morion skill, which teaches the agent the board and Ask Mo, installs from Settings, Skills. Model sources live on the same screen: the Claude Code or Codex login you already have, an OpenRouter key, or a local Ollama.
- One click for Claude Code, Codex, Cursor, Cline, Zed, Windsurf and Gemini Antigravity; any other MCP client takes one config entry
- Indexing, Ask Mo and Auto-code run on the subscription you are signed in to; no API key
- Any model through OpenRouter, with a spend cap per workspace; offline with Ollama
- Reinstall or remove an integration from the same screen; removing it deletes the entry it wrote
Decide per folder and per note what agents can see, change or delete.
Your working notes, your journal and your production checklist should not have the same agent access. Set four flags per folder (visible, create, update, delete) and three per note. A hidden folder is invisible to every client and to Mo: not in search, not in listings, not in the index.
- Reference folders: visible only, so agents quote but never rewrite
- One sensitive note opts out on its own, whatever its folder allows
- Master switches per tool category: turn Delete off and no client can delete anything
- Audit log names the actor on every write: you, Claude Code, Codex, Mo
| Actor | Calls | Last seen |
|---|---|---|
| mcp:claude-code | 1,204 | 2m ago |
| mcp:codex | 318 | 41m ago |
| mcp:cursor | 57 | Yesterday |
Mo indexes your notes so a ticket arrives with the context the agent needs.
Mo reads the notes you allow and writes a summary and keywords for each one, a catalog per folder and topic documents per project. When an agent picks up a ticket it calls one tool and gets the ticket, the comments that matter, the project brief, related decisions, lessons, specs and rules.
- Summary and keywords per note, editable; agents filter on them before they read the body
- Topics are created, merged and retired as the notebook changes; your own overview text stays
- Runs on the Claude Code login you already have, or the Codex login on macOS; no API key
- Indexes notes only, never your code or anything else on the disk
Ask Mo about the project and get an answer with the notes it came from.
Mo Assistant is a chat inside the app over your own notes and tasks. Every claim in an answer cites its note and carries a status: decided, provisional, proposed or superseded. Contradictions come back as conflicts, missing facts as gaps, never as a guess.
- Find a decision, a spec or a ticket without knowing where it is
- Set the app up, build an Auto-code workflow, create or review tickets from the chat
- Filter by project, folder and board status
- Try three questions on the Mo page
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?
Drag a ticket to Todo. Claude writes the code, Codex reviews it, you merge with one click.
Auto-code runs a ticket through a workflow of agents inside the app. The default one: Mo checks the ticket, Claude writes the diff, Mo decides, Codex reviews, done, with the diff on its own branch. Change the agent in any slot, add a stage, or draw your own graph with a gate that asks you in chat when the agent has a question and resumes when you answer.
- Five shipped templates, from “Code only” to “Plan + plan review + code + code review”
- Runs on your Claude Code or Codex subscription, or any model through OpenRouter, with a spend cap
- Trunk guard reverts changes that escape the worktree; AI merge resolver; you merge with one click, or let the folder auto-merge
- Ask your CLI agent or Mo to build the workflow for you over MCP
Connect the agent you already use.
Download Morion, connect Claude Code or Codex, and ask it to put the plan on the board.
Download