Skip to main content
from praisonaiagents import Agent

agent = Agent(
    name="profiled-agent",
    instructions="You are a customer support specialist.",
    profile="support",
)
agent.start("Help a customer reset their password.")
Use built-in agent profiles and modes to spin up specialised assistants without rewriting prompts from scratch. The user describes a support issue; the agent loads the right profile and responds in that mode.

​
How It Works

The agent loads the named profile’s prompt and tools, then answers the user in that mode.

​
Features

  • Built-in Profiles - Pre-configured agents for common tasks
  • Agent Modes - Primary, subagent, and all-context modes
  • Custom Profiles - Register your own agent profiles
  • Profile Discovery - List and filter available profiles

​
Built-in Profiles

ProfileModeDescription
generalPrimaryGeneral-purpose coding assistant
coderAllFocused code implementation
plannerSubagentTask planning and decomposition
reviewerSubagentCode review and quality
explorerSubagentCodebase exploration
debuggerSubagentDebugging and troubleshooting

​
Quick Start

1

Use a built-in profile

from praisonaiagents.agents.profiles import get_profile, list_profiles

coder = get_profile("coder")
print(f"Coder temperature: {coder.temperature}")

for profile in list_profiles():
    print(f"{profile.name}: {profile.description}")
2

Register a custom profile

from praisonaiagents.agents.profiles import register_profile, AgentProfile, AgentMode

security_auditor = AgentProfile(
    name="security_auditor",
    description="Security vulnerability scanner",
    mode=AgentMode.SUBAGENT,
    system_prompt="You are a security expert...",
    tools=["read_file", "search"],
    temperature=0.2,
    max_steps=40,
)
register_profile(security_auditor)

​
Agent Modes

ModeDescription
PRIMARYMain agent that can spawn subagents
SUBAGENTSpawned by another agent for specific tasks
ALLCan be used in any context

​
API Reference

​
AgentProfile

@dataclass
class AgentProfile:
    name: str                    # Unique profile name
    description: str             # What this agent does
    mode: AgentMode              # Execution mode
    system_prompt: str           # System prompt
    tools: List[str]             # Available tools
    temperature: float           # LLM temperature
    max_steps: int               # Maximum execution steps
    hidden: bool                 # Hide from listings

​
Functions

def get_profile(name: str) -> Optional[AgentProfile]:
    """Get a profile by name."""

def list_profiles(include_hidden: bool = False) -> List[AgentProfile]:
    """List all available profiles."""

def register_profile(profile: AgentProfile) -> None:
    """Register a custom profile."""

def get_profiles_by_mode(mode: AgentMode) -> List[AgentProfile]:
    """Get profiles that work in a specific mode."""

​
Examples

​
Using Built-in Profiles

from praisonaiagents.agents.profiles import get_profile

# Get the coder profile
coder = get_profile("coder")

# Use profile settings
print(f"System prompt: {coder.system_prompt[:100]}...")
print(f"Tools: {coder.tools}")
print(f"Temperature: {coder.temperature}")

​
Custom Profile

from praisonaiagents.agents.profiles import (
    register_profile,
    AgentProfile,
    AgentMode
)

# Create custom profile
security_auditor = AgentProfile(
    name="security_auditor",
    description="Security vulnerability scanner",
    mode=AgentMode.SUBAGENT,
    system_prompt="You are a security expert...",
    tools=["read_file", "search"],
    temperature=0.2,
    max_steps=40
)

register_profile(security_auditor)

​
Filter by Mode

from praisonaiagents.agents.profiles import (
    get_profiles_by_mode,
    AgentMode
)

# Get all subagent-capable profiles
subagents = get_profiles_by_mode(AgentMode.SUBAGENT)
for profile in subagents:
    print(f"{profile.name}: {profile.description}")

​
Best Practices

Use general, coder, or reviewer before authoring custom profiles — they already bundle sensible tools and temperature defaults.
Reserve SUBAGENT profiles for delegated tasks; use PRIMARY or ALL for agents that own the conversation.
One clear purpose per profile (e.g. security audit, planning) keeps tool lists small and prompts easier to maintain.
Set hidden=True on profiles still in development so they stay out of list_profiles() until ready.

Handoffs

Delegate work to subagents that can reuse profile presets.

Agent-Centric API

Consolidated feature flags and config objects on Agent(...).