Bots & Agents

A bot in AI-Parrot is a stateful conversational entity wrapped around an AbstractClient. An agent extends that with tool calling and a ReAct-style reasoning loop. A crew orchestrates several agents together via sequential, parallel or DAG-based execution.

What lives here

• AbstractBot — base class for everything conversational.
• Chatbot — single-LLM, single-turn-aware, stateful chat.
• Agent / BasicAgent — tool-using ReAct agents.
• AgentCrew — orchestrator with three execution modes:
• run_sequential() — output of agent n feeds agent n+1
• run_parallel() — agents run concurrently, results merged
• run_flow() — DAG defined with task_flow()
• Plugin agents — see parrot.agents for the dynamic-import registry that resolves agents by name.

Picking the right primitive

You need…	Use
A single-LLM conversation, no tools	Chatbot
One agent + a set of tools	Agent / BasicAgent
Several agents in a pipeline	AgentCrew.run_sequential
Independent agents fanning out	AgentCrew.run_parallel
Branching dependencies between agents	AgentCrew.run_flow

Read next

• Agents, Agent Mesh
• Crews, Crew Handler, Crew Summary
• Orchestration and Advanced Orchestration

API reference

API Reference → Bots