devduck 0.2.0

🦆 Extreme minimalist self-adapting AI agent - one file, self-healing, runtime dependencies

🦆 DevDuck

One file. Self-healing. Adaptive.

Minimalist AI agent that fixes itself when things break.

Install

pipx install devduck

Requires: Python 3.10+, Ollama running

Use

# Start DevDuck (auto-starts TCP, WebSocket, MCP servers)
devduck

# CLI mode
devduck "what's the time?"

# Python
import devduck
devduck("calculate 2+2")

Auto-Started Servers

When you run devduck, three servers start automatically:

• 🌐 Web UI: http://cagataycali.github.io/devduck (auto-connects)
• 🔌 TCP: nc localhost 9999 (raw socket)
• 🌊 WebSocket: ws://localhost:8080 (structured JSON)
• 🔗 MCP: http://localhost:8000/mcp (Model Context Protocol)

Connect via MCP

Add to your MCP client (e.g., Claude Desktop):

{
  "mcpServers": {
    "devduck": {
      "command": "uvx",
      "args": [
        "strands-mcp-server",
        "--upstream-url",
        "http://localhost:8000/mcp/"
      ],
      "disabled": false
    }
  }
}

Connect via Terminal

# Direct TCP connection
nc localhost 9999
> what's the time?

Features

• Self-healing - Auto-fixes deps, models, errors
• Hot-reload - Create tools in ./tools/*.py, use instantly
• Adaptive - Picks model based on OS (macOS: 1.7b, Linux: 30b)
• 14 tools - shell, editor, files, python, calculator, tcp, etc.
• History aware - Remembers shell/conversation context
• Multi-protocol - TCP, WebSocket, MCP, CLI, Python

Create Tool

# ./tools/greet.py
from strands import tool

@tool
def greet(name: str) -> str:
    return f"Hello {name}!"

Save. Done. Use immediately.

Multi-Model

export MODEL_PROVIDER="bedrock"
export STRANDS_MODEL_ID="us.anthropic.claude-sonnet-4-5-20250929-v1:0"
export STRANDS_ADDITIONAL_REQUEST_FIELDS='{"anthropic_beta": ["interleaved-thinking-2025-05-14", "context-1m-2025-08-07"], "thinking": {"type": "enabled", "budget_tokens": 2048}}'
export STRANDS_MAX_TOKENS="64000"

devduck "analyze data"

---

Quack. 🦆

Built with Strands Agents SDK