# Quickstart

**Start locally on your computer or remotely on your [fl0w cloud computer](/remote).** New here? Read the [Overview](/docs/overview) or [How it works](/docs/how-it-works) first.

## Starting fl0w on CLI (Recommended start)

### 1. Install fl0w by running this command in your terminal.

*macOS / Linux*

```bash
curl -fsSL https://raw.githubusercontent.com/jackmuva/fl0w/main/install.sh | bash
```

*Windows (PowerShell)*

```powershell
irm https://raw.githubusercontent.com/jackmuva/fl0w/main/install.ps1 | iex
```

### 2. Then login to fl0w. On your first time running this command, you will need to create an account. See [`fl0w login`](/docs/cli#account-and-sync) in the CLI reference for related commands.

```bash
fl0w login
```

### 3. From there, you can run fl0w directly in your terminal command line, and it will start the fl0w application.
```bash
fl0w
```

<Image src="/docs/tui.png" width={2010} height={1474}
  alt="Fl0w running side by side locally and in the cloud"
  className="mt-4 h-auto w-full max-w-2xl rounded-lg border border-line/80" />



## Starting fl0w remotely

After creating an account at [fl0w remote](/remote), you can use the fl0w agent in your cloud computer, just like the CLI application.

***Note:** If you started with the CLI, all of your MCPs, skills, and configurations are [automatically synced](/docs/how-it-works#zero-configs) with your cloud computer.*

<Image src="/docs/remote-tui.png" width={2402} height={1766}
  alt="Fl0w running side by side locally and in the cloud"
  className="mt-4 h-auto w-full max-w-2xl rounded-lg border border-line/80" />

## Configuring your agent

There are 3 main configurations you can choose at the start of your sessions: 

1. **Models**: *Certain models are more performant/cost-effective at certain tasks*
2. **MCP**: *Bring 3rd-party tools like Notion and Linear* — configure servers in [Adding MCP connectors](/docs/guides/mcp-connectors) or with [`fl0w mcp`](/docs/cli#mcp-servers).
3. **Skills**: *Choose from installed skills from your project, from Claude, from Cursor* — pick them per session with [`/configure`](/docs/slash-commands).

Fl0w's harness borrows concepts from [Opencode](https://opencode.ai) and [Claude Code](https://claude.com/product/claude-code) for the 
latest practices in tool search and skill loading.

Fl0w was designed to be **customizable and lightweight**. The configurations allow you to be intentional with your models, the context windows, and capabilities as needed.

## Generating and running your first workflow

Fl0w core use case is workflows that marry deterministic behavior with non-determinitic LLMs. 

To **templatize** the actions you took in a session, use [`/generate`](/docs/slash-commands) to turn any session into a workflow template. The [Automating sessions](/docs/guides/automating-sessions) guide covers triggers and cloud runs in depth.

<Image src="/docs/workflow.png" width={2318} height={1484}
  alt="Fl0w running side by side locally and in the cloud"
  className="mt-4 h-auto w-full max-w-2xl rounded-lg border border-line/80" />

To run the workflow with additional inputs, use the [`/workflow`](/docs/slash-commands) command (or [`fl0w workflow`](/docs/cli#fl0w-workflow) headlessly).

## Running workflows in the cloud

After you `/generate` a workflow, you can run these workflows locally with the `/workflow` command

OR run it remotely in your cloud computer. With cloud runs, you can run fl0w workflows:

1. Every week, day, or hour
2. Triggered via a webhook event
3. A form fill you can share with colleagues

<Image src="/docs/cron-trigger.png" width={1984} height={1698}
  alt="cron trigger"
  className="mt-4 h-auto w-full rounded-lg border border-line/80" />

Check your workflow steps and your run history directly in the **"Automate"** tab.

<Image src="/docs/workflow-history.png" width={2400} height={1560}
  alt="cron trigger"
  className="mt-4 h-auto w-full rounded-lg border border-line/80" />

## Using your cloud computer workspace

Your cloud computer has its own filesystem. You can use the filesystem to:

1. Create files and docs in markdown for easier reading
2. Ask fl0w to read files as a way to load context
3. Create folders to organize your work
4. Clone repos and upload files for specific work — see [Setting up codebases](/docs/guides/codebases)

<Image src="/docs/workspace.png" width={2338} height={1760}
  alt="cron trigger"
  className="mt-4 h-auto w-full rounded-lg border border-line/80" />


## Optimize performance and cost

As you use fl0w more and more workflows, track your agent's performance and cost to make optimizations. We'll be adding more features so you can tune your 
workflows for your use cases.

<Image src="/docs/optimize.png" width={2800} height={1764}
  alt="cron trigger"
  className="mt-4 h-auto w-full rounded-lg border border-line/80" />