Familiar logo
Familiar usefamiliar.com
Documentation / Fleet

Your First Agent

Create and interact with your first Fleet agent in under 5 minutes.

Prerequisites

Before starting, make sure you have:

  • Fleet installed (Installation Guide)
  • An AI provider configured (Anthropic, OpenAI, or another provider)

Step 1: Create a New Agent

Press Cmd+N (or click the + button) to create a new agent.

Fleet creates a new agent with default settings:

  • Name: "New Agent" (you can rename it)
  • Model: Your default model
  • Execution tier: Read-only

Step 2: Configure Your Agent

Rename the Agent

Click on the agent name in the header and type a new name, like "Dev Helper".

Select a Model

Click the model selector in the top-right to choose which AI model to use. Different models have different capabilities and costs.

Recommended for getting started: Claude Sonnet (good balance of capability and cost)

Set Permissions

Click the permissions icon to configure what the agent can do:

Tier What It Can Do
Restricted Safe read-only operations only
Read-Only Read any file, run safe commands
Read/Write Full file access, run any approved command

For your first agent, Read-Only is a safe choice for exploration.

Step 3: Send Your First Message

Type a message in the input field and press Enter:

What files are in this directory? Can you summarize them?

The agent will:

  1. Read the current directory listing
  2. Examine relevant files
  3. Provide a summary

Step 4: Approve Tool Calls

When an agent needs to use a tool (read a file, run a command, etc.), you'll see an approval request.

Approval queue

Review the request and:

  • Approve — Allow this specific tool call
  • Approve All — Allow this tool pattern for the session
  • Deny — Block this tool call

Tip: For trusted operations, approving a pattern (like "Read files in this directory") speeds up your workflow.

Step 5: View Workspace Files

As the agent works, it may read or create files. View these in the Workspace panel:

  1. Click the Files tab in the sidebar
  2. See all files the agent has accessed
  3. Click any file to preview its contents

Files the agent creates or modifies are saved to the agent's workspace directory.

Step 6: Pause and Resume

Need to step away? You can pause an agent mid-task:

  • Pause — Click the pause button or press Cmd+P
  • Resume — Click resume or send a new message

The agent remembers its context and continues where it left off.

Agent States

Watch the status indicator to understand what your agent is doing:

State Indicator Meaning
Idle Gray Waiting for input
Running Green pulse Processing your request
Paused Yellow Temporarily stopped
Waiting Blue Waiting for tool result
Awaiting Approval Orange Needs you to approve a tool
Completed Green check Task finished
Error Red Something went wrong

Try These Examples

Now that you have an agent running, try these prompts:

Explore a Codebase

Give me an overview of this codebase. What language is it?
What are the main components?

Get Help with Code

Explain what this function does: [paste code]

Find Something

Find all files that contain the word "config"

Summarize

Read the README and give me a 3-sentence summary

Customize the System Prompt

Want your agent to behave differently? Add a custom system prompt:

  1. Click the Settings icon on the agent
  2. Find Custom Prompt
  3. Add instructions like:
You are a code reviewer. Focus on:
- Security issues
- Performance problems
- Code readability

Be concise and actionable in your feedback.

Next Steps

Now that you've created your first agent: