โ† Back to Docs

๐Ÿฆ† DevDuck Run Anywhere

Generate installation & configuration for any platform

1. Install DevDuck
2. Configure Your Agent
๐Ÿ’ฌ
System Prompt Optional
Customize agent personality & behavior
โ–ผ
Define how the agent should behave. Supports GitHub sync via system_prompt tool.
๐Ÿค–
Model Provider Required
Choose your AI model backend
โ–ผ
Default model for selected provider
Required for API authentication
Maximum tokens for model response
Controls creativity (0=deterministic, 2=creative)
๐Ÿ”ง
Tools Required
Select agent capabilities
โ–ผ
๐Ÿฆ† DevDuck Core
โšก Strands Core
๐ŸŽฎ Strands Fun (Device Access)
Format: package:tool1,tool2;another_package:tool3
๐ŸŒ
Servers New
WebSocket, MCP, Zenoh P2P networking
โ–ผ
WebSocket
Real-time bidirectional communication
MCP Server
Model Context Protocol for Claude Desktop
Zenoh P2P
Auto-discovery mesh networking
WebSocket port (default: 8080)
MCP HTTP port (default: 8000)
Auto-discovers other DevDuck instances on your network
๐ŸŒ™
Ambient Mode New
Background thinking when idle
โ–ผ
Start ambient thinking after this idle time
Maximum ambient thinking iterations
๐Ÿ’ก Type ambient or auto in the REPL to toggle modes at runtime
๐ŸŽฌ
Session Recording New
Record & replay agent sessions
โ–ผ
๐Ÿ“ฝ๏ธ Use devduck --record to start with recording enabled, or type record in the REPL. Sessions are saved to /tmp/devduck/recordings/ and can be resumed with devduck --resume session.zip
๐Ÿ”Œ
External MCP Servers Optional
Connect to MCP tool servers
โ–ผ
Configure external MCP servers to load tools from
๐Ÿง 
Knowledge Base Optional
Automatic RAG with Bedrock KB
โ–ผ
Amazon Bedrock Knowledge Base ID for automatic RAG
๐Ÿ“ฆ
Dependencies Optional
Extra packages & setup commands
โ–ผ
Space-separated pip packages to install
Space-separated system packages (apt/brew/pkg)
Shell commands to run after installation

๐Ÿฆ† DevDuck - Extreme Minimalist Self-Adapting Agent

Docs ยท GitHub ยท Models ยท Tools