# Run your first project

This page takes you from an empty workspace to the loop the rest of Legevio is built on: a task goes to your AI, a result comes back, you accept it, and the next task receives it. [Connect your AI](/start/connect-your-ai/) first.

## 1. Create the project

On the empty Dashboard, press **Create your first project**. The wizard has four steps:

1. **Type**: pick the project type, such as **Game / Story**, **Book / Novel** or **Custom**. It is the only required choice; the name can wait.
2. **Team**: the roles the project starts with. The main roles are ticked. Every team is led by an AI **Project Manager**.
3. **Board**: the starter tasks, the collections for your Project Memory (for a story, for example, **Characters**) and the project documents.
4. **Inputs**: a project brief and any text files you already have. If you connected a computer, this step shows it.

Press **Open the project**. Creating a project does not start any AI work.

You now have five empty project documents (**Overview**, **Context**, **Decisions**, **Notes**, **Asset rights**), your roles and the starter tasks.

## 2. Write the project state page

On the Dashboard, find **Project state** and press **Start from template**. Fill in the sections: **Where we are**, **In progress**, **Next**, **Waiting on the owner**, **Risks** and **Working rules**. A few lines each is enough.

This is the page every chat and agent starts from. Your AI can keep it current itself, and every version is kept, so you can restore an earlier one from **History**. It is a working page, not Project Memory: nothing on it needs review.

## 3. Create a first task

Open **Tasks** and press **New task** (or the `C` key). Give it a **Title**, a **Brief** that says what done looks like, and under **Assign to** pick one of the project's AI roles from **Roles**. Press **Create task**.

A good first task asks for one decision you can check, for example "Propose the game's core loop in five sentences".

## 4. Let your AI do it

In the AI you connected, ask it to do the task by its number, for example: "Read Legevio task #1 and do it."

Your AI reads the task with its brief and context, does the work, and hands it in with one call. The result arrives in **Inbox → Drafts** and the task moves to **Review**. To hand in, your AI needs the **Create and update tasks** and **Create drafts** permissions; see [Connect your AI](/start/connect-your-ai/#use-a-personal-token).

If your AI is not connected, use **Copy brief** on the task, do the work anywhere, and paste the result back with **Submit result**.

## 5. Review and accept

Open **Inbox → Drafts** and the result. Read it, and the evidence your AI reported with it. Then choose:

- **Add to Project Memory** if the result is a decision the project should keep. The card shows where it will be filed, for example under **Decisions**; press **Change** to file it elsewhere.
- **Accept result** if the work is good but is not project truth. The task closes as Approved and Project Memory stays as it is.
- **Request revision** or **Reject** if it is not right yet.

## 6. See the next task receive it

Create a second task that builds on the first, open it, and choose its **Task Context** tab. The record you just accepted is listed under **Included in this brief**. The next chat, the next agent and the next person all start from it.

That is the whole loop: task → result → review → Project Memory → the next task.

## Set it up this way from day one

1. A project state page that every chat and agent starts from.
2. Review, then accept. Turn on an acceptance rule only where you want one.
3. The task number in every commit, and the commits reported with each hand-in, so every result shows the code it belongs to.
4. Every decision from chat becomes a Project Memory record.
5. Handoff files for procedures only. Status lives in Legevio.

## Next

- [Project Memory and acceptance](/concepts/project-memory/)
- [What the AI receives](/concepts/context/)
- [Inbox](/concepts/inbox/)