Agents are the core building blocks of the Agency Swarm framework. Each agent is specialized for a specific role and is designed to perform a specific set of processes within an agency.

​
Key Characteristics of Agents

Autonomous

Agents can determine the next best actions by themselves.

Adaptive

Agents adapt their course of action based on real-time feedback.

Interactive

Agents can manipulate their environment by using tools.

​
Agent Parameters

From a technical perspective, in Agency Swarm, agents are essentially wrappers for OpenAI Assistants. The Agent class includes convenience methods to help you manage the state of your assistant, upload files, attach tools, and more:
NameParameterDescription
ID (optional)idLoads the assistant from OpenAI assistant ID. Assistant will be created or loaded from settings if ID is not provided. Default: None
Name (optional)nameName of the agent. Default: Uses the class name
Description (optional)descriptionA brief description of the agent’s purpose. Default: None
Instructions (optional)instructionsPath to a file containing specific instructions for the agent. Default: Empty string
Tools (optional)toolsA list of tool classes that the agent can use (BaseTool, FileSearch, or CodeInterpreter). Default: None
Tool Resources (optional)tool_resourcesResources used by the assistant’s tools. For example, code_interpreter requires file IDs, while file_search requires vector store IDs. Default: None
Temperature (optional)temperatureControls randomness in the agent’s responses. Lower values make responses more focused and deterministic. Default: None
Top P (optional)top_pAlternative to temperature for controlling response randomness. Default: None
Response Format (optional)response_formatSpecifies the format for agent responses. Can be a string, dict, or Pydantic BaseModel. Default: "auto"
Tools Folder (optional)tools_folderPath to a directory containing tools. Each tool must be in a separate file named after the tool class. Default: None
Files Folder (optional)files_folderPath or list of paths to directories containing files for the agent. Default: None
Schemas Folder (optional)schemas_folderPath or list of paths to directories containing OpenAPI schemas. Default: None
API Headers (optional)api_headersHeaders for OpenAPI requests. Keys must match schema filenames. Default: Empty dict
API Params (optional)api_paramsExtra parameters for OpenAPI requests. Keys must match schema filenames. Default: Empty dict
Metadata (optional)metadataAdditional metadata for the agent. Default: Empty dict
Model (optional)modelThe OpenAI model to use. Default: "gpt-4o-2024-08-06"
Validation Attempts (optional)validation_attemptsNumber of attempts to validate responses. Default: 1
Max Prompt Tokens (optional)max_prompt_tokensMaximum tokens allowed in the prompt. Default: None
Max Completion Tokens (optional)max_completion_tokensMaximum tokens allowed in completions. Default: None
Truncation Strategy (optional)truncation_strategyStrategy for handling token limits. Default: None
Examples (optional)examplesList of example messages for the agent. Default: None
File Search (optional)file_searchConfiguration for the file search tool. Default: None
Parallel Tool Calls (optional)parallel_tool_callsWhether to run tools in parallel. Default: True
Refresh From ID (optional)refresh_from_idWhether to load and update the agent from OpenAI when an ID is provided. Default: True
Warning: The file_ids parameter is deprecated. Use the tool_resources parameter instead.

​
Agent Template

It’s recommended to create your agent in a seprate file. Your agent class should look like this: 
from agency_swarm import Agent

class AgentName(Agent):
    def __init__(self):
        super().__init__(
            name="agent_name",
            description="agent_description",
            instructions="./instructions.md",
            files_folder="./files",
            schemas_folder="./schemas",
            tools_folder="./tools",
            tools=[],
            temperature=0.3,
            max_prompt_tokens=25000,
            examples=[]
        )
You can add more parameters to the __init__ method. To initialize the agent: 
from AgentName import AgentName

agent = AgentName()