Install AI Workbench
Install the desktop app, set up your first project and team, and sign in to Mojo Up AI Cloud or your organisation.
In this tutorial you install AI Workbench, let it check your computer, create a project from a repository, start a first team of agents, and sign in. Allow fifteen minutes, most of it waiting for downloads.
You need nothing installed beforehand: the app checks the machine and installs what is missing, with a button for each thing.
1. Download and install
Get the Windows installer from the Download page and run it. It installs for your user only, so there is no administrator prompt. The installer is signed by Mojo Up.
AI Workbench is a native Windows app. Repositories can live on Windows or inside a WSL distro, where the agents then run in Linux.
The app keeps itself up to date. On Windows it downloads each update, checks Mojo Up's signature and installs it when you say so; on Linux an AppImage or .deb updates in place; on macOS the app shows the download link when an update is out.
2. Let it check this computer
Open AI Workbench. With no project yet, it offers to set one up and opens the wizard on This computer: a checklist of what the workbench needs here. Each row that is not satisfied has a button to install or fix it.
| Row | Why |
|---|---|
| Git | Required. Every agent works on its own branch in its own git worktree, which keeps agents out of each other's way. |
| Git identity | Your name and email. Agents commit as you, with the agent credited in a Co-authored-by: line. |
| Git Credential Manager | Optional. Signs you in when you clone or push over HTTPS; SSH keys work too. |
| WSL and a distro | Windows only, if you want agents to run inside Linux. Installing WSL needs a restart of Windows. |
| Docker | Only for repositories that live in a dev container. |
Choose Next when the rows are green, or when the ones left are ones you do not need.
3. Create your first project
A project is a name and one or more repositories that share a team, a task board, plans and memory.
Project
Give the project a name and a colour. The colour is how you recognise it in the project switcher.
Where
Choose which machine runs the agents: This computer, a WSL distro (Windows), a Dev container, or a computer Over SSH from your SSH config. Start with This computer if you are unsure. The wizard checks git on that machine, so for a WSL distro you may be asked for your name and email again.
Source
Choose where the code comes from:
- Clone a repository: paste an HTTPS or SSH URL from GitHub, Azure DevOps or any git host. If you have signed in under Settings → Integrations → Accounts, you can pick from your repositories instead.
- Existing folder: code that is already there. The wizard tells you what it found.
- New folder: made for you as a local git repository. It needs no remote.
If a clone stops because it could not sign in, choose Retry in the terminal, where git can ask for your credentials.
Prepare
The wizard clones or creates the repository and starts the workbench there. If the folder already carries workbench settings from someone else (a .mojoup/ folder), it tells you what comes with it: Adopt takes them, Start fresh moves them aside. Nothing is deleted either way.
Tools
The wizard checks the repository's machine for agent tools, such as Claude Code, GitHub Copilot CLI and Codex, with Install and Sign in beside each. You need at least one.
Team
Choose Start with a pair (a manager who plans and reviews, and one developer who builds), Feature squad (a manager, developers and a reviewer for larger pieces of work), or Later. You can change it any time in Teams.
Choose Finish. You now have a project, a running workbench and a team.
4. Give your agents some work
Open Chat for a conversation between you and one agent: the gentlest place to start. When you are ready for the team, open Team Work, describe a piece of work, and the manager plans it, hands out tasks, reviews what comes back and merges it. Tasks is the board those tasks live on; Source Control shows the branches the agents made and lets you land their work.
Nothing reaches your main branch behind your back: you merge the work, or open a pull request, when it is finished.
5. Sign in
AI Workbench works without an account. Sign in to share with a team, or to work in your organisation. Open Settings → Organisation.
- Choose Sign in with Mojo Up, then Continue with Mojo Up.
- Your browser opens on Mojo Up AI Cloud. Sign in with Google, Microsoft, GitHub, Facebook or a code emailed to you.
- Check the organisation the page names (your personal space, or one of your teams) and approve the workbench.
The workbench is now Free on your own, or Teams in a team. If you are new, see Sign in to Mojo Up AI Cloud and Create a team.
Each sign-in gets its own space, so your own projects and your organisation's stay apart. See Sign in and spaces.
Next steps
- Pair your phone to answer approvals away from your desk.
- Add a node to run agents on another machine.
- AI Workbench for everything the app does.