Core Concept: Tools, Memory, Plugins

These three components shape runtime behavior.

Tools

Allow the agent to act on external systems (files/APIs/etc).

Memory

Controls conversation state retention strategy.

Plugins

Inject policy, guardrails, retry, limits, and validation via lifecycle hooks.

Combined example

from nucleusiq.agents import Agent
from nucleusiq.agents.config import AgentConfig, ExecutionMode
from nucleusiq.memory import MemoryFactory, MemoryStrategy
from nucleusiq.plugins.builtin import ModelCallLimitPlugin, ToolRetryPlugin
from nucleusiq.prompts.zero_shot import ZeroShotPrompt
from nucleusiq.tools.builtin import FileSearchTool
from nucleusiq_openai import BaseOpenAI

agent = Agent(
    name="runtime-demo",
    prompt=ZeroShotPrompt().configure(system="You are a helpful assistant."),
    llm=BaseOpenAI(model_name="gpt-4.1-mini"),
    tools=[FileSearchTool(workspace_root="./data")],
    memory=MemoryFactory.create_memory(MemoryStrategy.SLIDING_WINDOW, window_size=10),
    plugins=[ModelCallLimitPlugin(max_calls=20), ToolRetryPlugin(max_retries=2)],
    config=AgentConfig(execution_mode=ExecutionMode.STANDARD),
)