Paid tier · BYO LLM key · part of the $8/mo bundle

Mo turns a ticket into a work packet.

Agents do better work when they start with the story around the card. Mo reads your local notes, tasks, comments, and activity, then gives Claude Code, Codex, Cursor, or Auto-code exactly the context they need: related tickets, decisions, risks, open questions, and recent changes. Runs on your own LLM key.

The big idea

The agent should not have to rediscover your project every time.

The deal has not changed: your notes, tasks, comments, and attachments live in one local file on your own machine. Mo adds an AI layer that works on that data — no cloud workspace, no copy-pasting, no "let me re-explain my project."

You point Mo at any LLM you already use: OpenRouter, OpenAI, Anthropic, Groq, or a local Ollama. You control the bill, you control where the prompt goes. Mo has three jobs:

  • Context broker — when Claude Code, Codex, or Cursor pick up a ticket, Mo packages linked notes, related tickets, decisions, open questions, risks, and recent folder activity.
  • In-app assistant — Ask Mo, fast lookup, drafts, plain-English workflow rules, Project Memory, approval gates before risky edits.
  • Decision node inside Auto-code workflows — every mo_stage asks Mo to decide: accept, reject, review, ask the human, or loop back.

Your external agents through MCP keep working the same way regardless. Mo just makes them faster, more accurate, and less likely to wander.

What Mo does

Six things Mo does on day one.

All of these are useful because Mo sees the board, the notebook, and the activity stream together. It is not a chatbot bolted onto a doc.

01 · Ask Mo

Ask anything about the work.

"What did Codex change in the trainer project this week?" "Which Auto-code tickets are stuck?" "What risks did we record around Lichess?"

Mo answers from your notes, your activity, your kanban — without you reading 40 cards. Sessions are saved like any other chat, you can come back to them.

02 · Fast lookup

Find the context behind the card.

"Where did we decide the move-validation rule?" "Which sibling tickets touch this API?" "What did the previous agent leave in Review?"

Mo runs hybrid keyword + semantic search across your entire workspace and pulls exactly what you meant — without you fighting search syntax.

03 · Watch the boards

Watches your boards, in plain English.

Tell Mo the rules for a folder once, in normal sentences:

"Claude can move cards from todo to review. Only I move anything to done. If a card sits in doing for more than two days, ask the owner what's blocking. Friday afternoon, draft a roll-up of what shipped this week."

Mo enforces those rules — nudges, catches drift, asks when unsure. You don't write code. You write what you want.

04 · Project Memory

A living summary of every folder.

Mo keeps a per-folder Project Memory that updates as work happens: what's been decided, what's stuck, what just shipped, what the conventions are.

Humans get a one-page status. Agents reading the folder pick up the up-to-date story instead of starting from scratch every session.

05 · Drafts & updates

Drafts the boring updates.

Friday spec roll-up. Status email. "What's changed since Tuesday." Investor update from the active boards. Optimisation suggestions when Mo notices stale or duplicated content.

Mo writes the draft from real activity, you edit and ship. The cadence is yours — on a schedule or on demand.

06 · Safety

Asks before anything risky.

Mo never deletes silently. Anything irreversible — delete a note, drop a folder, remove a tag — pops an approval card with the exact action and a one-line summary, and waits for your yes or no.

Edits, status moves, comments, drafts: those go through. Audit log records every action with the actor name "Mo", alongside Claude, Codex, Cursor, and you.

Mo's three roles

One assistant. Three jobs.

Mo doesn't try to be every AI in your workflow. Mo is the layer that knows your project, prepares context for the agents that do the work, and decides when to escalate to you.

1 · Assistant
Knows your project end to end
  • Ask Mo chat — grounded in your notes
  • Project Memory: per-folder living summary
  • Drafts: Friday roll-ups, status updates
  • Plain-English workflow rules
  • Approval gates for risky edits
2 · Context broker
Packages work packets for your agents

When Claude Code, Codex, Cursor, or Auto-code picks up a ticket, Mo does not hand them a one-line spec. Mo bundles:

  • Linked notes & related tickets
  • Decisions you made earlier in the project
  • Open questions, risks, blockers
  • Recent activity in the surrounding folder
  • Agents show up with the full picture
Auto-code · closed beta
3 · Workflow orchestrator
Decides between build stages

Every mo_stage inside an Auto-code workflow is Mo making a call after reading the ticket, the diff, the review, and the project context:

  • Diff written — accept, ask for review, or reject
  • Review came back — approve, reopen, or reject
  • Unsure? Pause and ask for your approval
  • Records the result, posts a closing comment
  • Same key as Mo's other roles — one LLM bill

All three roles run on the same LLM key — OpenRouter, OpenAI, Anthropic, Groq, or local Ollama. Point at Ollama and nothing leaves your machine.

Mo + your other agents

Mo briefs the agents. Your agents do the work.

Mo isn't competing with Claude, Codex, or Cursor. They have different jobs.

Mo
The librarian
  • Watches the whole project, all the time
  • Knows your workflow rules
  • Packages task-scoped context for the agent that picks up the ticket
  • Files agent-produced notes back into the right folders
  • Asks for your call when unsure
Your MCP agents
The doers
  • Pick up cards from your kanban (Claude Code, Codex, Cursor, Antigravity, Cline, Zed)
  • Write actual code, prose, designs
  • Run in their native client
  • Talk to your codebase, your IDE
  • Report back via Morion's audit log

A typical loop: you describe a task, Mo turns the surrounding project memory into a work packet, an agent works the card, Mo evaluates the result, and the board records what happened. You stay in the loop without staring at the loop.

Coming next on Mo

What lands as we go.

Paid subscribers get every new Mo capability as it ships — at the early-adopter price.

Auto-code · next
Approval checkpoints in Auto-code workflows.

Auto-code pauses when the workflow needs your call; Mo opens a chat with "here's the plan — go / no-go?" and your reply unblocks the next stage. After this lands, free open beta opens to everyone.

Soon
Structured questions back to you.

Mo asks structured "Ask Human" questions when stuck — choices, free text, deadlines. You answer in one click. Threaded into the activity panel, not buried in chat.

Soon
Open a card in your favourite agent.

From any kanban card: "Open in Claude Code", "Open in Codex", "Open in Cursor", "Copy work prompt." The card becomes the dispatch point.

Roadmap
Global Mo memory across projects.

Mo remembers your preferences, naming conventions, and decisions across folders — not just per-project. Asks before storing anything controversial; you can edit memory anytime.

Early-adopter pricing

Mo + Auto-code in the paid tier. $8/mo.

$8/mo billed annually ($96/yr), or $16/mo monthly. Bring any LLM key — OpenRouter, OpenAI, Anthropic, Groq, or local Ollama. Cancel anytime.

Single paid tier — everything we charge for, today and tomorrow, in one bundle.

FAQ

The questions you'd ask anyway.

Where does Mo run?

Mo runs on the LLM provider you choose — OpenRouter, OpenAI, Anthropic, Groq, or a local Ollama. You bring the key, you pay your provider, you control where prompts go. Your data lives on your disk; Mo only sees the slice it needs for the current task.

Can Mo run fully offline?

Yes — point Mo at a local Ollama (or any OpenAI-compatible local endpoint). Mo treats it the same as cloud providers. Nothing leaves your machine. This is as offline as Mo gets.

What does Mo see?

Only what's relevant to the current job. If you ask Mo about your API folder, Mo reads recent activity and notes from that folder. Mo never holds your whole notebook. Folders you mark hidden from AI are invisible to Mo as well — same gates as MCP.

Can Mo delete things?

Not without your explicit approval. Anything irreversible pops an approval card showing the exact action; Mo waits for your yes or no. Reversible actions — edits, status moves, comments, drafts — go through directly, with audit log entries.

What's the spending control?

You control the bill — your LLM provider, your account, your budget caps. Point Mo at OpenRouter and you can cap spend at the provider level. Point at Ollama and the bill is your electricity.

Do I have to use Mo on the paid tier?

No. Mo is opt-in per folder. You decide which projects Mo watches, which you keep manual, and which you hand entirely to your other AI clients via MCP. The paid tier just unlocks the option.

How does Mo compare to ChatGPT or Claude?

ChatGPT and Claude are general-purpose assistants. Mo is a specialised workspace assistant with built-in knowledge of your folders, kanban, activity, and rules. Mo is short on world knowledge and long on your project. Use Mo for "what's happening in my workspace" and use the others for "explain this concept" or "write this code."

If I cancel, what happens?

Your notes, boards, Project Memory entries, and Auto-code workflows stay — they live in your local data file, untouched. Mo stops watching new activity, Auto-code stops running. Re-subscribe and both pick up where you left off.

Try Morion Free. Add Mo + Auto-code when you're ready.

Free is the whole notebook with full MCP. Paid ($8/mo billed annually) adds Mo, Auto-code, and unlimited boards — at the early-adopter rate.