AgentCore Deploy¶

Deploy DevDuck to Amazon Bedrock AgentCore with one command.

Quick Deploy¶

devduck deploy --launch

This:

Creates a deployment directory with handler + requirements
Configures the agent with agentcore CLI
Launches the agent on AgentCore
Returns the agent ARN for invocation

Prerequisites¶

pip install bedrock-agentcore
aws configure  # or have valid AWS credentials

Commands¶

Deploy¶

# Configure only (no launch)
devduck deploy

# Configure AND launch
devduck deploy --launch

# Custom configuration
devduck deploy \
    --name code-reviewer \
    --tools "strands_tools:shell,file_read;devduck.tools:use_github,editor" \
    --model "us.anthropic.claude-sonnet-4-20250514-v1:0" \
    --system-prompt "You are a senior code reviewer" \
    --region us-west-2 \
    --launch

# Full example with all options
devduck deploy \
    --name my-agent \
    --tools "strands_tools:shell" \
    --idle-timeout 1800 \
    --max-lifetime 43200 \
    --no-memory \
    --no-otel \
    --env "MY_VAR=value" \
    --env "OTHER=123" \
    --force-rebuild \
    --launch

List Agents¶

devduck list
devduck list --region us-east-1

Check Status¶

devduck status --name my-agent

Invoke¶

devduck invoke "analyze this code" --name my-agent
devduck invoke "hello" --agent-id abc123

Deploy Options¶

Flag	Default	Description
`--name`	`devduck`	Agent name
`--tools`	default set	Tool configuration
`--model`	claude-sonnet-4	Model ID
`--region`	`us-west-2`	AWS region
`--launch`	false	Auto-launch after configure
`--system-prompt`	—	Custom system prompt
`--idle-timeout`	900	Idle timeout (seconds)
`--max-lifetime`	28800	Max lifetime (seconds)
`--no-memory`	false	Disable AgentCore memory
`--no-otel`	false	Disable OpenTelemetry
`--env`	—	Custom env vars (repeatable)
`--force-rebuild`	false	Force rebuild deps

Unified Mesh Proxy¶

DevDuck auto-starts an AgentCore proxy on ws://localhost:10000 that connects:

CLI DevDuck (local terminal)
Browser clients (mesh.html)
Deployed AgentCore agents (cloud)

All visible as peers in the same mesh:

# List all peers (local + cloud)
agentcore_proxy(action="status")

# Invoke a cloud agent from CLI
agentcore_invoke(agent_id="abc123", prompt="hello from CLI")