> ## Documentation Index
> Fetch the complete documentation index at: https://docs.semilattice.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Using the app

> Set up your context, start a project, and act on what comes back.

The app is at [app.semilattice.ai](https://app.semilattice.ai). Everything below happens inside a project, with the agent doing the planning for you.

![A project page with three simulations running in a round and the agent's conversation on the right](https://semilattice.ai/static/marketing/blog/projects/hero.png)

**1. Start a project**

Go to **Projects** and click **New project**.

**2. Say what you want to find out**

A new project opens with two cards. **Define a simulation** sets one up yourself. **Upload a document or images** gives the agent a plan, a spec, an ad or a screenshot to plan from. You can also paste a link or tell the agent what you want to find out.

The agent asks before it starts if anything is unclear.

![The agent planning two simulations from one request, both shown as Planning in the round](https://semilattice.ai/static/marketing/blog/projects/agent-plans.png)

**3. Set up your context**

The first time you plan something, the agent asks about your company or product and writes your context from your answers. It is the background every project reads, so it is not about this project. You can edit it any time under **Contexts**. See [How Semilattice works](/how-it-works#your-context).

**4. Check the plan**

The agent plans a set of simulations and picks a [user model](/user-models) for each. Open a simulation to read its questions or its task, and ask the agent to change anything before it runs. Nothing runs until you say so.

**5. Read the results**

Simulations take from a few minutes to around half an hour. When they finish, the [synthesis](/how-it-works#results-and-the-synthesis) for the round says what to change and why. Open any simulation to see its own results.

**6. Change your work and go again**

Apply what you learned, then **Start a new round** with the next version. Each round keeps its own simulations and synthesis, so you can see what changed between versions.

To watch a whole project go from a first round to the next version, see [Introduction to user-in-the-loop](https://semilattice.ai/user-in-the-loop/introduction), the first lesson in the guide.

## Elsewhere in the app

* **Simulations** lists every simulation you have run, across all projects.
* **User Models** lists the user models you can use, with each one's accuracy.
* **Contexts** holds your context and any others you can use.
* **Settings** has your workspace, email notifications and **API Keys**, for the [API](/api-reference/introduction).

## Share and export

Click **Share** on a simulation or a synthesis to make it viewable by anyone with the link. A synthesis can also be exported as a PDF report or Markdown.

## Carry on from your own agent

Projects you start in the app can be picked up from Claude, Claude Code, Cursor or any other MCP client, and projects your agent starts appear here. See [Semilattice MCP](/mcp/introduction).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.