This documentation is outdated. The platform is currently being upgraded to the latest version of ACP with major changes. Updated docs are expected in ~2 weeks. Apologies for the inconvenience.

Agent composition allows you to chain multiple specialized agents together into powerful workflows, enhancing their capabilities and enabling complex task execution.

By default, each agent receives the output from the previous agent as input. You can customize this behavior in the instructions by specifying how each agent should use the previous outputs.

​
GUI

The compose playground provides a visual interface for creating and running multi-agent compositions. Today, we have sequential composition which is powered by the sequential-workflow agent.

1

Launch the web interface

beeai ui
2

Navigate to the compose playground

3

Compose agents

  • Add agents to your workflow by selecting them from the available list.
  • Arrange them in the desired execution order.
  • For each agent, specify the instructions that describe what it should do.
4

Run your workflow

Click the Run button to execute the composition.

​
CLI

You can also compose agents directly from your terminal using the sequential-workflow agent, which orchestrates the execution of multiple agents in sequence.

There are two ways to work within the CLI:

​
Interactive mode

Interactive mode walks you through the process of creating a composition step by step:

beeai run sequential-workflow

The CLI will guide you through:

  1. Selecting the first agent to run
  2. Providing instructions for that agent
  3. Choosing whether to add another agent to the sequence
  4. Repeating steps 1-3 until your workflow is complete

​
Direct JSON input

For repeatable workflows, you can define your composition as a JSON structure:

beeai run sequential-workflow '{
  "steps": [
    {"agent": "first-agent", "instruction": "First step instructions"},
    {"agent": "second-agent", "instruction": "Second step instructions"}
  ]
}'

Replace first-agent and second-agent with available agents from the platform, and define instructions for each. You may add additional steps as needed.

To list all available agents, run: beeai list.