Skip to main content
Permission modes provide predefined permission behaviors for agents, controlling how they handle permission requests during execution.

​
Quick Start

1

Import Permission Mode

from praisonaiagents.permissions import PermissionMode
2

Use with Subagents

from praisonaiagents.tools.subagent_tool import create_subagent_tool

# Create subagent tool with permission mode
tool = create_subagent_tool(
    default_permission_mode="plan"  # Read-only mode
)

# Spawn subagent
func = tool["function"]
result = func(task="Explore the codebase")

​
Available Modes

ModeValueSafetyDescription
DEFAULTdefault✅ SafeStandard permission checking - prompts for each operation
ACCEPT_EDITSaccept_edits⚠️ ModerateAuto-accepts file edit operations without prompting
DONT_ASKdont_ask⚠️ ModerateAuto-denies all permission prompts
BYPASSbypass_permissions🔴 DangerousSkips all permission checks entirely
PLANplan✅ SafeRead-only exploration mode - no modifications allowed

​
Mode Details

The default mode prompts for permission on each operation that requires approval.
from praisonaiagents.permissions import PermissionMode

mode = PermissionMode.DEFAULT
print(mode.value)  # "default"
Behavior:
  • Prompts user for each permission request
  • Follows configured permission rules
  • Safest mode for production use
Automatically accepts file edit operations without user confirmation.
from praisonaiagents.permissions import PermissionMode

mode = PermissionMode.ACCEPT_EDITS
print(mode.value)  # "accept_edits"
Behavior:
  • Auto-approves file write operations
  • Still prompts for other operations
  • Useful for automated refactoring tasks
Use with caution - agents can modify files without confirmation.
Automatically denies all permission prompts without user interaction.
from praisonaiagents.permissions import PermissionMode

mode = PermissionMode.DONT_ASK
print(mode.value)  # "dont_ask"
Behavior:
  • Auto-denies all permission requests
  • Agent continues with read-only operations
  • Useful for safe exploration tasks
Bypasses all permission checks entirely.
from praisonaiagents.permissions import PermissionMode

mode = PermissionMode.BYPASS
print(mode.value)  # "bypass_permissions"
Behavior:
  • No permission checks performed
  • All operations allowed
  • Maximum agent autonomy
Security Risk: Only use in fully trusted environments. Agent can perform any operation without restriction.
Restricts agent to read-only operations for safe exploration.
from praisonaiagents.permissions import PermissionMode

mode = PermissionMode.PLAN
print(mode.value)  # "plan"
Behavior:
  • Only read operations allowed
  • Write/modify operations blocked
  • Perfect for codebase exploration
Recommended for subagents that only need to analyze code without making changes.

​
Usage with Subagents

​
Default Mode for All Subagents

from praisonaiagents.tools.subagent_tool import create_subagent_tool

# All subagents use plan mode by default
tool = create_subagent_tool(
    default_permission_mode="plan"
)

​
Per-Call Override

func = tool["function"]

# This call uses plan mode (read-only)
result1 = func(
    task="Explore the auth module",
    permission_mode="plan"
)

# This call uses accept_edits mode
result2 = func(
    task="Refactor the utils module",
    permission_mode="accept_edits"
)

​
Combined with Model Selection

tool = create_subagent_tool(
    default_llm="gpt-4o-mini",
    default_permission_mode="plan"
)

func = tool["function"]
result = func(
    task="Analyze code patterns",
    llm="gpt-4o",  # Override model
    permission_mode="default"  # Override permission mode
)

​
Best Practices

When subagents only need to read and analyze code, use plan mode to prevent accidental modifications.
The bypass_permissions mode should only be used in controlled development environments, never in production.
Choose the permission mode that matches the task requirements:
  • Exploration tasks → plan
  • Refactoring tasks → accept_edits
  • Interactive tasks → default
When using non-default modes, log the permission mode being used for audit purposes.

Permissions Module

Pattern-based permission rules

Subagent Delegation

Spawn and manage subagents