agent-utilities 0.2.20

Agent Utilities for Pydantic AI Agents

Agent Utilities - Pydantic AI Utilities

Version: 0.2.20

Overview

Agent Utilities provides a robust foundation for building production-ready Pydantic AI Agents. It simplifies agent creation, adds multi-agent supervisor patterns, and provides essential "operating system" tools for agents, including workspace management, scheduling, and discovery.

Key Features

• Agent Creation: Streamlined create_agent function that handles MCP servers, skills, and model configuration automatically.
• Multi-Agent Support: Native support for the supervisor pattern, allowing complex tasks to be delegated to specialized child agents.
• Agent Server: Built-in FastAPI server (create_agent_server) with SSE support for easy integration into web UIs and A2A networks.
• Elicitation Support: Built-in flow for handling structured user input requests from MCP servers, seamlessly integrated with the Web UI.
• Workspace Management: Automated management of agent state through standard markdown files (IDENTITY.md, MEMORY.md, USER.md).
• A2A Integration: Seamless discovery and communication between agents in a distributed network.
• Periodic Scheduler: In-memory task scheduler for running background agent jobs.
• Lightweight & Lazy: Core utilities are lightweight. Heavy dependencies like FastAPI or LlamaIndex are lazy-loaded only when requested via optional extras.

Installation

# Core utilities only
pip install agent-utilities

# With full agent support (recommended)
pip install agent-utilities[agent]

# With MCP server support
pip install agent-utilities[mcp]

# With embedding/vector support
pip install agent-utilities[embeddings]

Quick Start

from agent_utilities import create_agent

# Create a simple agent with workspace tools
agent = create_agent(name="MyAgent")

# Or create a multi-agent supervisor
agent = create_agent(
    name="Supervisor",
    agent_definitions=[{"name": "Researcher", "description": "Search the web"}]
)