Pipecat Subagents Overview

New to Pipecat Subagents? Check out the introduction and quickstart first.

Pipecat Subagents is a multi-agent framework for Pipecat that lets you build AI applications with multiple coordinated agents. Agents communicate through a shared bus, can hand off to each other, and coordinate work through structured task protocols.

Installation

uv add pipecat-ai-subagents

Optional extras for specific features:

# Pipecat Flows integration
uv add "pipecat-ai-subagents[flows]"

# Redis distributed bus
uv add "pipecat-ai-subagents[redis]"

# WebSocket proxy agents
uv add "pipecat-ai-subagents[websocket]"

Reference Pages

Agents

BaseAgent

Core agent class with lifecycle, bus communication, and task coordination

LLMAgent

Agent with LLM pipeline and automatic tool registration

FlowsAgent

Agent with Pipecat Flows integration for structured conversations

Infrastructure

AgentRunner

Lifecycle orchestrator for multi-agent systems

Bus

Bus infrastructure for inter-agent messaging

Messages

All bus message types for agent communication

Decorators & Types

Decorators

@tool, @task, and @agent_ready decorators

Types

Shared types, enums, and data classes

Exceptions

Error types for task coordination

Distributed

Proxy Agents

WebSocket proxy agents for distributed deployments

Serializers

Message serialization for network transport

Agent Type Hierarchy

Class	Parent	Description
`BaseAgent`	-	Core agent with lifecycle, bus, and task coordination
`LLMAgent`	`BaseAgent`	Adds LLM pipeline and `@tool` registration
`FlowsAgent`	`BaseAgent`	Adds Pipecat Flows integration
`WebSocketProxyClientAgent`	`BaseAgent`	Forwards bus messages to a remote server
`WebSocketProxyServerAgent`	`BaseAgent`	Receives bus messages from a remote client

Pipecat Server