A memory layer for you and AI

I want AI to handle my work “just right” more often.
But how?

Organize the knowledge, methods, and history behind your work, so you and AI can work from the same context. nohmitaina (noh for short) gives you a way to delegate work to AI without relying solely on instructions for each task.

Work with AI agents on macOS and Windows

02 · Watch the video

See how to delegate work to AI with noh.

Organize tasks, knowledge, and procedures in noh so AI can refer to what you have organized. An LP update shows the flow from preparation and execution to reviewing the results.

  1. Review the background and approach to the work
  2. Let AI work with the information it needs
  3. Keep the results to inform the next task

Music: Inspired by Kevin MacLeod, CC BY 4.0. Edited and mixed with narration. Full credit.

Optional captions appear below the video in five languages. The final Episode was recorded separately from the preceding CLI run.

03 · Does this sound familiar?

“Vague instructions lead to vague results.”

You keep giving instructions and checking the work to get closer to the result you want. What is missing is not a better prompt, but a way to keep sharing the context and methods behind your work.

A · REPEAT

Repeating instructions and corrections every time

You explain the required references, caveats, and output format again. If a correction stays in that one conversation, you will repeat it on the next task.

B · MAINTAIN

Building a way to delegate work to AI takes effort

You have to design the steps, roles, decisions, inputs, outputs, and constraints yourself, then maintain them as AI and your work change.

C · OVERVIEW

The setup gets complex, and you lose the big picture

Even when you save knowledge and procedures, unclear relationships make it time-consuming to check what to reference and where to make changes.

Share the work itself with AI. That is what is needed. But building a system for it is hard.
04 · What is noh?

noh lets you and AI review and develop “the work itself” together in one place. It is a “brain-like” memory editor.

What do we mean by “the work itself”? Let’s take a landing page update as an example.

05 · How noh works

Represent your work with four memory types.

In noh, work is organized in an Area (noh) hierarchy. Here, “Product → Marketing → Landing page update (example)” shows how work in a selected Area is organized into four memory types.

ProductMarketingLanding page update (example)
CONCEPT

What you know, and why you think that way

Organize the guidelines, background, and reasoning you repeatedly refer to when updating a landing page in a tree.

├─ Landing page update guidelines
│ ├─ Target audience
│ └─ Writing conventions
└─ Value to communicate
PROCEDURE

How to proceed

For each step, organize who checks what and where a person makes the decision.

Review messaging guidelinesMarketing role
Draft the copyMarketing role
Human reviewHuman
Apply to the landing pageMarketing role
TASK

What to do now

Manage current work and link it to the Concept and Procedure it needs.

Example: Add information next to the document request button explaining how to receive the document and what happens after requesting it
EPISODE

What actually happened

Record actions, results, and observations to improve your knowledge and methods for next time.

Example: A pre-publication review revealed that it was unclear how to receive the document. We added the steps after a request and included this in the next review checklist
PERSONAMarketing role (Persona)

This role handles steps in the landing page update. It defines what to check from a marketing perspective at each step of the Procedure.

06 · How these challenges change

Turn detailed instructions for every task into a way of working you can develop.

Here is how noh changes the burdens in A–C.

A · REPEAT — Repeating instructions and corrections every time

Move reusable knowledge out of individual instructions.

Organize knowledge in a Concept tree and reference what you need from a Task. Feed earlier observations from Episode back into your knowledge and methods so the next Task can reference them too.

CONCEPT + TASK + EPISODEREFERENCE
Concept tree
▾ Build a landing page
 ├ Target audience
 ├ Value to communicate
 └ Writing conventions
Task · Landing page structure and copy

Reference only the knowledge this task needs.

Target audienceValue to communicateWriting conventions
Episode: Previous correction: “Lead headings with the conclusion” → Apply to Concept “Writing conventions” → Reference in the next Task
B · MAINTAIN — Designing and maintaining the setup becomes another job

Review AI procedures visually and revise them step by step.

Procedure provides a structure and UI for organizing steps, Persona, references, inputs, outputs, and checks. It reduces the effort of designing the fields and their presentation from scratch. See AI-handled steps and human decisions in one flow, and revise only the steps that need it.

Actual screen showing the 10 steps in noh-dev

This demo illustrates the interaction and differs from the actual app screen. It reproduces the ten stages of the screenshot version of noh-dev using fixed data, with shortened details. Added examples are marked “Sample”. Editing, saving and running tools are not available.

The interactive demo on this page works on desktop computers only.

C · OVERVIEW — Files multiply, making the big picture and changes hard to track

Move between the big picture and the details, from areas of work to individual steps.

Separate work into Areas and review the relationships between knowledge in the Concept tree. Procedure shows both the overall flow and the roles, inputs, and outputs of a selected step. Save current results and observations in Episode and connect them to the knowledge or steps to review next.

AREA + CONCEPT + PROCEDURE + EPISODECONTEXT MAP
AreaMarketingDevelopmentResearch
Concept tree · Marketing
▾ Brand guidelines
 ├ Target audience
 ├ Value to communicate
 └ Writing conventions
PROCEDURE · Update the landing page

Decide the structure → Write the copy → Review the wording → Publish
Selected step: Review Persona, references, inputs, and outputs

EPISODE · TODAY

Replaced the Area explanation with a concrete example. Refer to the same writing conventions in the next landing page update.

07 · Get started

Start organizing Areas with the built-in area-advisor Persona.

Tell area-advisor about your goals and everyday work, review its suggested Area hierarchy, and choose one Area to start with.

AI agents read and write information through noh CLI. They need an environment that can run noh CLI and access the storage location. For Claude Code and Codex, noh provides a way to install a Skill that explains how to use it.

area-advisor is included when a new workspace is created. Updating the app does not add it to existing workspaces, including those created in 0.4.0 or earlier. If needed, use the English Persona text below.
Read area-advisor (full English text)

noh is free. Depending on the AI service you use, separate charges may apply.

Organize Areas with area-advisorDiscuss your goals and everyday work with the built-in AI Persona, then review its suggested hierarchy
Choose your first AreaExample: Product > Marketing > Landing page update
What does it contain?Think of your current Tasks, background knowledge, and frequently used references
What work do you repeat often?Make the work you do in the same order each time your first Procedure
Learn from the resultsAfter execution, feed the observations recorded in Episode back into Concept and Procedure, and revise them.
08 · Happy Customer Wanted!!

Share your work with AI.

Bring your tasks, methods, knowledge, and results into one shared memory for you and AI. Help us improve noh: tell us about the work you want to tackle and what you notice when using it.

“How could I set up this workflow in noh?”
“I tried it, and this part was difficult to use.”
“I want to save these repeated instructions for next time.”
OUR PROMISE

We will keep improving how you share work with AI.

AI models will continue to evolve. To put their capabilities to use in real work, you need to be able to share your knowledge, methods, and expected outcomes with AI.

That is what we are building with noh. Organize tasks, methods, related knowledge, and results so you can understand and review the work while delegating it to AI.

To make AI truly useful at work, we promise to keep working in this field and improving noh.

Choose a download

What shipped, and when.

VersionDateHighlights
0.4.12026-11-15Adds Today boards and Task groups, inherited Task status labels for Areas, and improvements to memory links, reference notes, and Persona management. Improves the Agent CLI, imports, and desktop reliability.
0.4.02026-10-24Adds Areas for organizing memories and improves Concept editing. Notes now default to Memo (save only) and inherit workspace and folder settings. Fixes Windows upgrade backups and startup.
0.3.02026-10-01Adds Tasks, Personas, Procedures, and Episodes, with improved editing and navigation.
0.2.12026-08-14Adds Windows support, native menus, inline editing in Summon, and a startup guide in four languages. Improves editing and navigation responsiveness, makes saving notes and quitting the app safer, and improves the reliability of LLM settings.
0.2.02026-06-25The noh agent CLI: AI assistants can read and write your noh. Activation visualizations (sparkline and barometer) and stability fixes.
0.1.72026-06-12Link cards resolve their titles automatically. Hardened security (CSP, SSRF guard on URL clips) and third-party notices. Editor sync and June usability fixes.
0.1.62026-05-20Editor document tabs with macOS titlebar integration. Local Whisper model settings.
0.1.52026-05-18Manual Recall from selected text. Rectangular selection and macro record/replay. Markdown rendering fixes. More resilient Dream processing.
0.1.42026-05-11Multi-workspace activation. Startup catch-up for missed Dream runs. Scoring and Core refinements.