Jot: say what you need.
AI helpers do the work.
You stay in charge.
Jot turns a normal conversation into a tracked work order. You approve it. Your own computer does the work with AI helpers, and the results come back to you.
Talk
Speak to ChatGPT as you would to an assistant.
Approve
Each new job waits for your OK on your phone.
Get results
A report, files and pictures, ready to review.
From a sentence to finished work, in six steps
The voice part listens and talks. It does not do the work.
What the voice assistant does
- Turns your speech into text and replies out loud.
- Works out what you want, and asks when it is not sure.
- Can read the board for you: “What is still running?”
- Fills in Jot's job form and sends it.
What it does not do
- It does not run code or programs.
- It does not touch your computer or your files.
- It cannot approve its own jobs.
- It does not wait for the work to finish. The work starts later.
Jot's own phone app also has a microphone button. It turns a voice note into text for your notes. That is transcription only. It does not start work.
ChatGPT fills in a work order through Jot's connector
Jot gives ChatGPT a small set of tools: buttons that it can press for you. The standard for these buttons is called MCP (Model Context Protocol). Think of a waiter who writes your order on a ticket and puts it on the kitchen rail.
What happens, in order
- ChatGPT first reads Jot's current rules and the jobs already on the board.
- It picks the kind of work: research, writing, code, routine or social.
- It sends the job. Jot replies with a job number.
- The conversation can end here. The job lives on the board.
Other tools let it read a job, list the board, add a correction to a job, or ask for a rerun. You must sign in to connect ChatGPT to Jot.
Example only. The job numbers and text are made up.
The job board keeps every request, with its status
The board is a cloud service. It stores each job, its brief, its progress and its results. You see the same board in the phone app, on the desktop and in chat.
One source of truth
Chat, phone and desktop all read and write the same list.
Nothing gets lost
If the PC is off or busy, jobs wait safely on the board.
A clear history
Who asked, when, what changed, and what the helper said.
Nothing starts without your OK, unless you choose to trust it
The default: approve each job
Every job that an assistant adds waits on the board until you approve it in the Jot phone app. The assistant cannot approve or start it.
Trust, when you want speed
You can trust one sign-in (for example, your own ChatGPT). Then jobs that you asked for in that chat can start without a tap, within limits that you set: how many at once, a pause, an end date.
The runner on Moya gives each job to the right AI helper
Moya is the owner's own PC (a gaming PC). A program on it, the runner, watches the board, takes approved jobs and starts an AI helper for each one.
| Kind of job | Default helper |
|---|---|
| code | Claude (Anthropic) |
| research | Codex (OpenAI) |
| writing | Gemini (Google) |
| social | Gemini (Google) |
| routine | Claude Sonnet |
The owner can change these defaults in the apps.
Its own workspace
Each job gets its own folder. A code job gets its own copy of the project on its own branch, so jobs do not overwrite each other.
Time and budget limits
Each run stops after a fixed time limit (4 hours today). Usage is recorded, and the PC only starts a job when it has room for it.
No surprises on screen
Helpers run in the background, with no pop-up windows. Browsers run hidden.
You can follow the work, then read the results
Live progress
Short updates on what the helper does, as it works.
A plain report
What it did, what it checked and what is still open.
A clear outcome
Complete or incomplete, with a reason. Unfinished work is never shown as done.
Files and pictures
Screenshots, videos and documents appear on the board during the work.
Not quite right? Correct it and go again.
is on record
Add a correction
Tell the assistant “change the tone” or “we decided X”. Jot keeps it beside the job as an amendment, with who said it and when.
Rerun or continue
Start the same job again from the beginning, or let a new job continue where the last one stopped.
Shared notes
Decisions and facts go into notes in Jot. The assistant can attach a note to a later job, so the next job starts better informed.
Jobs know their project, and can wait for each other
Project context
A project is a body of work, such as an app, with its own folder on Moya and a main copy. Jobs with no code use “none”. Shared rules for helpers go into each brief.
Dependencies
A job can wait for up to 10 other jobs. It stays blocked until all of them are done. If one fails, it stays blocked and says why.
Example: market research → post ideas
Two research jobs run side by side. One writing job waits for both and combines them.
This pattern comes from Jot's own rules for assistants.
Build on earlier work
A job that waits for one job in the same project starts from that job's work, or waits until that work is merged.
Code changes are checked before they join the main copy
Own branch
Each code job works on a private copy of the project. The main copy does not change while it works.
Checks and review
The helper runs the project's tests. A merge job can bring the branch up to date and run the tests again.
You decide
Each project has a merge rule: automatic after checks, review or manual. New projects start as manual.
Layers of control, not blind trust
Approval gate
New jobs wait for your OK. A chat cannot approve jobs, change project settings or unpause the board.
Sign-in for every tool
ChatGPT must sign in to Jot before it can use any tool. Moya and the board talk only over an encrypted connection.
Fenced workspaces
Each helper works only in its own job folder or project copy, with temporary files kept apart.
Careful merges
A merge needs your request, or the project's automatic rule plus passing checks. It is never forced. Helpers must not push or deploy without your approval.
Limits
A time limit per job, a limit on jobs at once, a memory check before each start, and optional weekly allowances per project.
Secrets stay out
Activity logs are shortened and cleaned of secrets. Helpers are told to keep passwords and keys out of reports.
What this means for a small business
Real kinds of requests
- writing “Write five hooks for the next video.”
- social “What do people say online about this kind of product?”
- research “Review this change and tell me what is risky.”
- code “Fix the sign-in bug and test it.”
- routine “Run the tests and send me a report.”
These examples come from Jot's own rules for assistants.
Ideas do not wait for the desk
Give work as soon as you think of it, in plain speech.
Several specialists, one place
Use the best AI for each kind of work, without learning each tool.
Control and a paper trail
You approve, you review, and you can see every step later.
How the parts connect (technical names)
Glossary
MCP: Model Context Protocol, a standard way for an AI assistant to use outside tools. Tool call: one structured request from the assistant to a tool.
Runner: the program on Moya that takes jobs and starts helpers. Engine: the AI helper program (Claude, Codex, Gemini).
Branch: a private copy of a project's code. Merge: accept that copy into the main copy. Amendment: a correction kept beside a job.