Quick Start

Get BeeSystem running locally in under 5 minutes. This guide installs Bee Worker, creates your first agent, and runs it.

Prerequisites

Requires macOS, Linux, or Windows with Docker. Node.js 18+ for the SDK.

1. Install Bee Worker

Install the Bee Worker runtime on your machine:

macOS / Linuxbash
curl -fsSL https://get.beesystem.ai/worker | sh
Windows (Scoop)bash
scoop install bee-worker
Dockerbash
docker pull beesystem/bee-worker:latest

Verify the installation:

bee --version

2. Create a new agent project

Use the bee init command to scaffold a new agent:

bee init my-first-agent
cd my-first-agent

This creates the following structure:

Project structuretext
my-first-agent/
  agent.yaml        # Agent specification
  tools/            # Tool definitions
  memory/           # Memory configuration
  tests/            # Agent test scenarios

3. Inspect the agent spec

Open agent.yaml to see the default agent configuration:

agent.yamlyaml
name: my-first-agent
version: "0.1.0"
description: A starter BeeSystem agent

capabilities:
  - reason
  - use_tools
  - remember_context

tools:
  - name: web_search
    type: builtin
    config:
      max_results: 10

memory:
  type: session
  max_tokens: 8192

runtime:
  timeout: 30s
  max_retries: 2

4. Start Bee Worker

Start the local Bee Worker runtime:

bee worker start

5. Run your agent

Run the agent with a task:

bee agent run \
  --spec ./agent.yaml \
  --task "Summarize the key principles of distributed systems"

You should see structured output with the agent's reasoning, tool calls, and final response.

6. Install the SDK (optional)

For programmatic agent control, install the Bee SDK:

npm install @beesystem/sdk
index.tstypescript
import { BeeAgent } from "@beesystem/sdk";

const agent = BeeAgent.fromSpec("./agent.yaml");

const result = await agent.run({
  task: "Summarize the key principles of distributed systems",
});

console.log(result.output);

You're up and running

You've created and run your first BeeSystem agent. Next, read Core Concepts to understand agents, tools, and the runtime model.

Next steps