strands

Agent Control + Strands Integration

High-level integration guide for adding Agent Control safety guardrails to AWS Strands agents using plugins and steering. This keeps your agent code unchanged while enforcing centralized policies.

Overview

Agent Control integrates with Strands using two native extension points:

• AgentControlPlugin for deny-style enforcement at model and tool stages
• AgentControlSteeringHandler for steer-style guidance using Strands Guide()

Use the plugin for hard blocks and the steering handler for soft guidance. You can run both together.

Installation

pip install agent-control-sdk[strands-agents]

Integration Patterns

Plugin pattern (hard blocks)

Use the plugin to enforce controls before and after model calls, tool calls, node transitions, and invocation.

from agent_control.integrations.strands import AgentControlPlugin
from strands import Agent
import agent_control

agent_control.init(agent_name="my-agent")

plugin = AgentControlPlugin(
    agent_name="my-agent",
    event_control_list=[BeforeToolCallEvent, AfterToolCallEvent],
    enable_logging=True,
)

agent = Agent(
    name="my_agent",
    model=model,
    tools=[...],
    plugins=[plugin],
)

Steering pattern (soft guidance)

Use the steering handler to convert steer actions into Guide() instructions for the next LLM call.

from agent_control.integrations.strands import AgentControlSteeringHandler
from strands import Agent
import agent_control

agent_control.init(agent_name="banking-agent")

steering_handler = AgentControlSteeringHandler(
    agent_name="banking-agent",
    enable_logging=True,
)

agent = Agent(
    name="banking_agent",
    model=model,
    tools=[...],
    plugins=[steering_handler],
)

What this enables

• Multi-stage protection across model calls, tool calls, and node transitions
• Centralized policy updates without redeploying agents
• Hard blocks and soft guidance depending on control action
• Zero code changes to existing Strands tools and workflows

Error handling

• ControlViolationError for hard blocks (deny)
• ControlSteerError for steer actions (plugin only)

from agent_control import ControlViolationError

try:
    result = await agent.invoke_async("Send email with SSN 123-45-6789")
except ControlViolationError as exc:
    print(f"Blocked by: {exc.control_name}")

Docs and Examples

Read more about this integration and follow through examples:

• Docs: https://docs.agentcontrol.dev/integrations/strands
• Examples: ../../../../../../examples/strands_agents/