devduck 1.10.0

🦆 Self-adapting agent - one file.

🦆 DevDuck

Self-modifying AI agent that hot-reloads its own code—builds itself as it runs.

One Python file that adapts to your environment, fixes itself, and expands capabilities at runtime. Deploy anywhere — terminal, browser, cloud, or all at once in a unified mesh.

DevDuck landing screen — your agent, ready to quack.

Your browser does not support the video tag. Watch the intro video.
DevDuck in 90 seconds ▶

Learn more: https://dev.duck.nyc

🎬 See It In Action

Feature	What You'll See	Demo
🔥 Hot-Reload	Agent detects code changes and restarts instantly	Watch
🖥️ TUI Mode	Multi-conversation Textual UI with streaming markdown	devduck --tui
🌐 Web UI	Clean web interface with real-time streaming	Watch
🛠️ Dynamic Tools	Save .py file in ./tools/ → use instantly	Watch
🌊 TCP Streaming	Connect via netcat, apps, or other agents	Watch
🔗 Zenoh P2P	Auto-discover & coordinate multiple DevDucks	Multi-terminal magic ✨
🎬 Session Recording	Record, replay & resume agent sessions	Time-travel debugging 🕰️
🎬 Asciinema Recording	Record agent trajectories as .cast files	Play anywhere with asciinema play 🎥
🌙 Ambient Mode	Background thinking while you're idle	Auto-explores topics 🧠
🔌 IPC & Tray	macOS menu bar + Unix socket IPC
💬 Ambient Overlay	Floating AI input with glassmorphism UI	Watch
🌐 Unified Mesh	Connect CLI + browser + cloud agents in one mesh	All agents, one network 🕸️
☁️ Deploy CLI	devduck deploy --launch to AgentCore	One-command cloud deploy 🚀
🧩 Browser Peers	Browser tabs join the mesh as first-class peers	Open mesh.html to join 🌍
🍎 macOS Control	Calendar, Mail, Safari, Finder, System & more	One tool for your Mac 💻
🎵 Spotify	Full playback, playlists, discovery control	Music while you code 🎶
💬 Messaging	Telegram, WhatsApp, Slack auto-reply bots	Multi-platform chat 📱
🎮 RL/ML Toolkit	Train RL agents, fine-tune LLMs with LoRA	Machine learning built-in 🧠

---

Quick Start

# Install & run
pipx install devduck && devduck

# Multi-conversation TUI (concurrent panels, streaming markdown)
devduck --tui

# One-shot query
devduck "create a REST API with FastAPI"

# Python API
python -c "import devduck; devduck('analyze this code')"

# Session recording (time-travel debugging)
devduck --record "analyze this codebase"
# → Exports to /tmp/devduck/recordings/session-*.zip

# Asciinema recording (shareable agent trajectories)
DEVDUCK_ASCIINEMA=true devduck
# → Records to /tmp/devduck/casts/devduck-*.cast
# → Play with: asciinema play /tmp/devduck/casts/devduck-*.cast

# Resume from recorded session
devduck --resume session-20250202-123456.zip "continue where we left off"

# Deploy to AWS AgentCore
devduck deploy --launch
devduck deploy --name my-agent --tools "strands_tools:shell,editor" --launch

Requirements: Python 3.10-3.13, AWS credentials (or Ollama/Anthropic/OpenAI/GitHub/Gemini/MLX)

---

Core Capabilities

Feature	What It Does	How to Use
🔥 Hot-Reload	Auto-restarts on code changes	Edit __init__.py → saves → auto-restart
🖥️ TUI Mode	Multi-conversation concurrent terminal UI	devduck --tui
🛠️ Runtime Tools	Add/remove tools without restart	manage_tools(action="add", ...)
📦 Dynamic Loading	Install packages and load tools on-the-fly	install_tools(action="install_and_load", package="...")
🧠 Auto-RAG	Remembers conversations via Knowledge Base	Set DEVDUCK_KNOWLEDGE_BASE_ID
🧠 SQLite Memory	Persistent local memory with full-text search	sqlite_memory(action="store", content="...")
🌊 Multi-Protocol	TCP, WebSocket, MCP, IPC servers	Auto-starts on ports 10001, 10002, 10003
🔗 Zenoh P2P	Auto-discover & coordinate with other DevDucks	zenoh_peer(action="broadcast", message="...")
🌐 Unified Mesh	Connect CLI + browser + cloud agents	Auto-starts relay on port 10000
☁️ Deploy CLI	One-command AgentCore deployment	devduck deploy --launch
🔌 MCP Client	Connect to external MCP servers	Set MCP_SERVERS env var
🎬 Session Recording	Record & replay entire sessions	devduck --record or session_recorder()
🎥 Asciinema Recording	Record terminal output as .cast files	DEVDUCK_ASCIINEMA=true devduck
💾 State Time-Travel	Save/restore agent state	state_manager(action="export")
🌙 Ambient Mode	Background thinking when idle	DEVDUCK_AMBIENT_MODE=true or type ambient
⏰ Job Scheduler	Cron and one-time scheduled agent tasks	scheduler(action="add", name="...", schedule="...")
📋 Background Tasks	Parallel agent tasks in separate threads	tasks(action="create", prompt="...")
💬 Interactive Dialogs	Rich terminal UI dialogs (forms, pickers)	dialog(dialog_type="input", text="...")
🔔 Notifications	Native OS notifications (macOS, terminal, TUI)	notify(title="Done", message="Task complete")
🖥️ Computer Control	Mouse, keyboard, screenshots, scrolling	use_computer(action="screenshot")
🎤 Speech Listener	Background Whisper transcription	listen(action="start")
🔍 LSP Integration	Real-time code diagnostics from language servers	lsp(action="diagnostics", file_path="...")
📝 AGENTS.md	Auto-loads project context from AGENTS.md	Place AGENTS.md in working directory
📝 Self-Improvement	Updates own system prompt	system_prompt(action="add_context", ...)
☁️ AWS Deploy	One-command serverless	devduck deploy --launch
🍎 macOS Control	Calendar, Mail, Safari, Finder, System, Keychain & more	use_mac(action="calendar.events")
📝 Apple Notes	Create, edit, search, export notes	apple_notes(action="list")
🎵 Spotify	Full playback, search, playlists, discovery	use_spotify(action="now_playing")
💬 Telegram	Bot listener + full API access	telegram(action="start_listener")
💬 WhatsApp	Messaging via wacli (no Cloud API)	whatsapp(action="send_text", to="...", text="...")
💬 Slack	Socket Mode listener + messaging	slack(action="start_listener")
🔌 JSON-RPC	Generic RPC client (HTTP & WebSocket)	jsonrpc(method="getInfo", endpoint="...")
🎮 RL/ML Toolkit	Train RL agents, fine-tune LLMs	rl(action="train", env_id="CartPole-v1")
🎤 Speech-to-Speech	Real-time voice conversations	Nova Sonic, OpenAI Realtime, Gemini Live
🧩 Apple Silicon	On-device NLP, Vision OCR, sensors, WiFi, SMC	apple_nlp(action="sentiment", text="...")

---

Developer Setup

git clone [e-mail]:cagataycali/devduck.git
cd devduck
python3.13 -m venv .venv
source .venv/bin/activate

# Install
.venv/bin/pip3.13 install -e .

devduck

---

Architecture

graph TB
    A[User Input] -->|CLI/TUI/TCP/WS/MCP/IPC| B[DevDuck Core]
    B -->|Auto RAG| C[Knowledge Base]
    C -.->|Context Retrieval| B
    B -->|Tool Calls| D[60+ Built-in Tools]
    D --> E[shell/editor/calculator]
    D --> F[GitHub/AgentCore]
    D --> G[TCP/WebSocket/MCP/IPC]
    D --> H[tray/ambient/dialog/notify]
    D --> O[Telegram/WhatsApp/Slack]
    D --> P[use_mac/apple_notes/spotify]
    D --> Q[rl/jsonrpc/scraper]
    D --> R[scheduler/tasks/sqlite_memory]
    D --> S[listen/lsp/use_computer]
    B -->|Hot-reload| I[./tools/*.py + __init__.py]
    I -.->|Load Instantly| D
    B -->|Runtime| K[manage_tools/install_tools]
    K -.->|Expand| D
    B <-->|Zenoh P2P| L[Other DevDucks]
    L -.->|Auto-discover| B
    B <-->|Unified Mesh| M[Browser Peers + AgentCore]
    M -.->|Ring Context| B
    B -->|Deploy| N[devduck deploy → AgentCore]
    B -->|AGENTS.md| T[Project Context Auto-Load]
    B -->|Response| J[User Output]
    J -.->|Store Memory| C
    
    style B fill:#e1f5ff
    style C fill:#d4edda
    style I fill:#fff3cd
    style K fill:#ffe6cc
    style L fill:#f0e6ff
    style M fill:#ffe6f0

Self-adapting loop: Query → RAG → Tools → Response → Memory → Hot-reload/Runtime-load → Repeat

---

TUI Mode (Multi-Conversation)

Launch a full Textual-based terminal UI with concurrent interleaved conversations:

devduck --tui

Features:

• Multiple conversations running in parallel with color-coded panels
• Streaming markdown rendering (tables, code blocks, lists)
• Tool call tracking with icons and timing
• Collapsible panels — click header to toggle
• Zenoh peer sidebar with live updates
• Slash commands: /clear, /status, /peers, /tools, /help
• Keyboard shortcuts: Ctrl+L (clear), Ctrl+K (focus), Ctrl+T (toggle sidebar)
• Agent can push content to the TUI via the tui() tool

# Agent-side: push rich content to TUI
tui(action="panel", content="## Analysis\nResults here...", title="Report")
tui(action="notify", content="Task complete!", style="green")
tui(action="status", content="Processing 3/10 files...")

---

Model Setup

DevDuck auto-detects providers based on credentials:

Priority: Bedrock → Anthropic → OpenAI → GitHub → Gemini → Cohere → Writer → Mistral → LiteLLM → LlamaAPI → MLX → Ollama

Provider	API Key	Auto-Detected
Bedrock	AWS credentials	✅ If boto3 auth succeeds
Anthropic	ANTHROPIC_API_KEY	✅ If key present
OpenAI	OPENAI_API_KEY	✅ If key present
GitHub	GITHUB_TOKEN or PAT_TOKEN	✅ If key present
Gemini	GOOGLE_API_KEY or GEMINI_API_KEY	✅ If key present
Cohere	COHERE_API_KEY	✅ If key present
Writer	WRITER_API_KEY	✅ If key present
Mistral	MISTRAL_API_KEY	✅ If key present
LiteLLM	LITELLM_API_KEY	✅ If key present
LlamaAPI	LLAMAAPI_API_KEY	✅ If key present
MLX	No key needed	✅ On Apple Silicon (M1/M2/M3/M4)
Ollama	No key needed	✅ Fallback if nothing else found

Just set your API key - DevDuck handles the rest:

export ANTHROPIC_API_KEY=sk-ant-...
devduck  # Auto-uses Anthropic

export OPENAI_API_KEY=sk-...
devduck  # Auto-uses OpenAI

export GOOGLE_API_KEY=...
devduck  # Auto-uses Gemini

Manual override:

export MODEL_PROVIDER=bedrock
export STRANDS_MODEL_ID=us.anthropic.claude-sonnet-4-20250514-v1:0
devduck

---

Tool Management

Runtime Tool Management

Add, remove, or reload tools while agent is running:

# List all loaded tools
manage_tools(action="list")

# Add tools from a package at runtime
manage_tools(action="add", tools="strands_fun_tools.cursor,strands_fun_tools.clipboard")

# Create a tool on the fly
manage_tools(action="create", code='''
from strands import tool

@tool
def greet(name: str) -> str:
    """Greet someone by name."""
    return f"Hello, {name}!"
''')

# Fetch and load a tool from GitHub
manage_tools(action="fetch", url="https://github.com/user/repo/blob/main/my_tool.py")

# Discover tools in a package
manage_tools(action="discover", tools="strands_tools", verbose=True)

Dynamic Package Installation

Install Python packages and load their tools at runtime:

# Discover available tools before loading
install_tools(action="list_available", package="strands-fun-tools", module="strands_fun_tools")

# Install package and load all tools
install_tools(action="install_and_load", package="strands-agents-tools", module="strands_tools")

# Install and load specific tools only
install_tools(
    action="install_and_load",
    package="strands-fun-tools", 
    module="strands_fun_tools",
    tool_names=["clipboard", "cursor", "bluetooth"]
)

# Load tools from already installed package
install_tools(action="load", module="strands_tools", tool_names=["shell", "calculator"])

# List currently loaded tools
install_tools(action="list_loaded")

Static Tool Configuration

Format: package1:tool1,tool2;package2:tool3,tool4

# Minimal (shell + editor only)
export DEVDUCK_TOOLS="strands_tools:shell,editor"

# Dev essentials
export DEVDUCK_TOOLS="strands_tools:shell,editor,file_read,file_write,calculator"

# Full stack + GitHub
export DEVDUCK_TOOLS="devduck.tools:tcp,websocket,mcp_server,use_github;strands_tools:shell,editor,file_read"

devduck

Hot-Reload Tools from Directory

Create ./tools/weather.py:

from strands import tool
import requests

@tool
def weather(city: str) -> str:
    """Get weather for a city."""
    r = requests.get(f"https://wttr.in/{city}?format=%C+%t")
    return r.text

Enable directory auto-loading:

export DEVDUCK_LOAD_TOOLS_FROM_DIR=true
devduck
# Save weather.py → use instantly (no restart needed)

Default: Directory loading is OFF. Use manage_tools() or install_tools() for explicit control.

---

New Tools

⏰ Job Scheduler

Cron-based and one-time job scheduling with disk persistence:

# Start the scheduler daemon
scheduler(action="start")

# Add a cron job (every 6 hours)
scheduler(action="add", name="backup", schedule="0 */6 * * *", prompt="Run backup check")

# Add a one-time job
scheduler(action="add", name="remind", run_at="2026-03-04T15:00:00", prompt="Remind about meeting")

# List all jobs
scheduler(action="list")

# View execution history
scheduler(action="history", name="backup")

# Run a job immediately
scheduler(action="run_now", name="backup")

📋 Background Tasks

Run parallel agent tasks in separate threads:

# Create a background task
tasks(action="create", task_id="research", prompt="Research quantum computing", system_prompt="You are a researcher")

# Check status
tasks(action="status", task_id="research")

# Send follow-up messages
tasks(action="add_message", task_id="research", message="Focus on healthcare applications")

# Get result
tasks(action="get_result", task_id="research")

# List all tasks
tasks(action="list")

🧠 SQLite Memory

Persistent local memory with full-text search, tagging, and metadata:

# Store a memory
sqlite_memory(action="store", content="Important finding about X", title="Research", tags=["research", "x"])

# Full-text search
sqlite_memory(action="search", query="important finding")

# List recent memories
sqlite_memory(action="list", limit=5)

# Raw SQL access
sqlite_memory(action="sql", sql_query="SELECT COUNT(*) FROM memories")

# Export
sqlite_memory(action="export", export_format="json")

# Stats
sqlite_memory(action="stats")

💬 Interactive Dialogs

Rich terminal UI dialogs using prompt_toolkit:

# Text input
dialog(dialog_type="input", text="What's your name?", title="Name")

# Yes/No confirmation
dialog(dialog_type="yes_no", text="Deploy to production?", title="Confirm")

# Radio selection
dialog(dialog_type="radio", text="Pick a color:", options=[["red", "Red"], ["blue", "Blue"]])

# Multi-select checkbox
dialog(dialog_type="checkbox", text="Select features:", options=[["auth", "Auth"], ["api", "API"]])

# Multi-field form
dialog(dialog_type="form", text="User details:", form_fields=[
    {"name": "email", "label": "Email"},
    {"name": "role", "label": "Role", "default": "developer"}
])

# File picker
dialog(dialog_type="file", text="Select a file:", path_filter="*.py")

# Password input
dialog(dialog_type="password", text="Enter password:")

🔔 Notifications

Native notifications across platforms:

# macOS notification (uses terminal-notifier or osascript)
notify(title="Build Complete", message="All tests passed!", sound="default")

# With URL action
notify(title="PR Merged", message="Click to view", url="https://github.com/...")

# Speech notification
notify(title="Alert", message="Deployment started", method="speech")

🖥️ Computer Control

Mouse, keyboard, screenshots, and app switching:

# Take screenshot
use_computer(action="screenshot")

# Click at position
use_computer(action="click", x=500, y=300)

# Type text
use_computer(action="type", text="Hello world")

# Keyboard shortcuts
use_computer(action="hotkey", keys=["cmd", "c"])

# Scroll
use_computer(action="scroll", direction="down", clicks=5)

# Drag
use_computer(action="drag", x=100, y=100, to_x=500, to_y=500)

# Switch app
use_computer(action="switch_app", app_name="Safari")

# Get screen size
use_computer(action="screen_size")

🎤 Background Speech Listener

Background audio capture with Whisper transcription:

# Start listening (uses default microphone)
listen(action="start", model_name="base")

# Start with trigger keyword
listen(action="start", trigger_keyword="hey duck")

# Start in auto mode (triggers on long speech)
listen(action="start", auto_mode=True, length_threshold=50)

# Check status
listen(action="status")

# Get transcripts
listen(action="get_transcripts", limit=10)

# List audio devices
listen(action="list_devices")

# Stop
listen(action="stop")

🔍 LSP Integration

Real-time code diagnostics from language servers (pyright, typescript-language-server, gopls, rust-analyzer, clangd):

# Get diagnostics for a file
lsp(action="diagnostics", file_path="main.py")

# Go to definition
lsp(action="definition", file_path="main.py", line=10, character=5)

# Find references
lsp(action="references", file_path="main.py", line=10, character=5)

# Hover documentation
lsp(action="hover", file_path="main.py", line=10, character=5)

# Document symbols
lsp(action="symbols", file_path="main.py")

# Check status
lsp(action="status")

Auto-diagnostics hook (opt-in): Automatically appends LSP diagnostics after file-modifying tool calls:

export DEVDUCK_LSP_AUTO_DIAGNOSTICS=true
devduck
# Now every file edit automatically shows type errors/warnings

📝 Conversation Management

View, drop, compact, and export agent conversation history:

# List all turns
manage_messages(action="list")

# List only user messages
manage_messages(action="list", role="user")

# List all tool calls with IDs
manage_messages(action="list_tools")

# Drop specific turns
manage_messages(action="drop", turns="0,2")

# Compact turns (strip tool blocks, keep text)
manage_messages(action="compact", start=0, end=5)

# Export conversation
manage_messages(action="export", path="/tmp/chat.json")

# Import conversation
manage_messages(action="import", path="/tmp/chat.json")

# Get stats
manage_messages(action="stats")

# Clear all
manage_messages(action="clear")

---

Apple Silicon Tools

On-device AI tools that run entirely on Apple's Neural Engine — zero cloud calls:

🧠 Apple NLP

# Language detection
apple_nlp(action="detect", text="Bonjour le monde")

# Sentiment analysis
apple_nlp(action="sentiment", text="This product is amazing!")

# Named entity recognition
apple_nlp(action="entities", text="Tim Cook announced new MacBook in Cupertino")

# Part-of-speech tagging
apple_nlp(action="pos", text="The quick brown fox jumps")

# Word embeddings & similarity
apple_nlp(action="distance", word="king", word2="queen")
apple_nlp(action="similar", word="computer", top_k=10)

# Lemmatization
apple_nlp(action="lemma", text="The dogs were running quickly")

👁️ Apple Vision

# OCR from image
apple_vision(action="ocr", image_path="/path/to/screenshot.png")

# OCR the current screen
apple_vision(action="ocr_screen")

# Detect barcodes/QR codes
apple_vision(action="barcode", image_path="/path/to/image.png")

# Detect faces
apple_vision(action="faces", image_path="/path/to/photo.jpg")

# Detect rectangles (documents, cards)
apple_vision(action="rectangles", image_path="/path/to/scan.png")

📡 Apple WiFi

# Current connection details
apple_wifi(action="status")

# Scan nearby networks
apple_wifi(action="scan")

# Signal quality analysis
apple_wifi(action="signal")

# Best channel recommendation
apple_wifi(action="best_channel")

# Full diagnostics
apple_wifi(action="diagnostics")

🌡️ Apple Sensors & SMC

# Full sensor status
apple_sensors(action="status")

# Temperature readings
apple_smc(action="temps")

# Fan speeds
apple_smc(action="fans")

# Power draw
apple_smc(action="power")

# Battery details
apple_sensors(action="battery")

# Keyboard backlight
apple_sensors(action="set_keyboard", brightness=0.5)

---

AGENTS.md Project Context

Place an AGENTS.md file in your working directory to automatically inject project-specific context into every agent query:

# AGENTS.md

## Project: My API Service
- Framework: FastAPI
- Database: PostgreSQL
- Tests: pytest + httpx
- Deploy: Docker on ECS

## Conventions
- Use async/await everywhere
- Type hints required
- Tests in tests/ directory

DevDuck auto-loads this file on startup and includes it in the system prompt. No configuration needed.

---

Context Window Auto-Recovery

DevDuck automatically handles context window overflow:

1. When the conversation history exceeds the model's context limit, DevDuck detects the error
2. It clears the message history automatically
3. Retries the latest query with a fresh context
4. No manual intervention needed — the agent self-heals

This prevents the agent from getting stuck on long-running sessions.

---

Speech-to-Speech

Real-time voice conversations with multiple providers:

# Start speech session with Nova Sonic (AWS Bedrock)
speech_to_speech(action="start", provider="novasonic")

# Start with OpenAI Realtime API
speech_to_speech(action="start", provider="openai")

# Start with Gemini Live
speech_to_speech(action="start", provider="gemini_live")

# Custom voice and settings
speech_to_speech(
    action="start",
    provider="novasonic",
    model_settings={
        "provider_config": {"audio": {"voice": "matthew"}},
        "client_config": {"region": "us-east-1"}
    }
)

# Stop session
speech_to_speech(action="stop", session_id="speech_20250126_140000")

# Check status
speech_to_speech(action="status")

# List conversation histories
speech_to_speech(action="list_history")

# List available audio devices
speech_to_speech(action="list_audio_devices")

Supported Providers:

• Nova Sonic (AWS Bedrock): 11 voices (English, French, Italian, German, Spanish)
• OpenAI Realtime API: GPT-4o Realtime models
• Gemini Live: Native audio streaming

Features:

• Background execution (parent agent stays responsive)
• Tool inheritance from parent agent
• Conversation history saved automatically
• Natural interruption with VAD
• Custom audio device selection

---

MCP Integration

As MCP Server (Expose DevDuck)

Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "devduck": {
      "command": "uvx",
      "args": ["devduck", "--mcp"]
    }
  }
}

Or start HTTP MCP server:

mcp_server(action="start", port=8000, stateless=True)
# Connect at: http://localhost:8000/mcp

Modes: --mcp (stdio for Claude Desktop) | http (background server) | stateless=True (multi-node)

As MCP Client (Load External Servers)

Expand capabilities by loading tools from external MCP servers:

export MCP_SERVERS='{
  "mcpServers": {
    "strands-docs": {"command": "uvx", "args": ["strands-agents-mcp-server"]},
    "remote": {"url": "https://api.example.com/mcp", "headers": {"Auth": "Bearer token"}},
    "custom": {"command": "python", "args": ["my_server.py"]}
  }
}'
devduck

Supported transports: stdio (command/args/env) | HTTP (url/headers) | SSE (url with /sse path)

Tool prefixing: Each server's tools get prefixed (e.g., strands-docs_search_docs)

---

Zenoh Peer-to-Peer Networking

Auto-discover and coordinate multiple DevDuck instances across terminals or networks.

How It Works

1. Each DevDuck joins a Zenoh peer network
2. Multicast scouting (224.0.0.224:7446) auto-discovers peers on local network
3. Peers exchange heartbeats to maintain presence awareness
4. Commands can be broadcast to ALL peers or sent to specific peers
5. Responses stream back in real-time

Quick Start

# Terminal 1: Start DevDuck (Zenoh enabled by default)
devduck
# 🦆 ✓ Zenoh peer: hostname-abc123

# Terminal 2: Start another DevDuck
devduck
# 🦆 ✓ Zenoh peer: hostname-def456
# Auto-discovers Terminal 1!

# Terminal 1: See discovered peers
🦆 zenoh_peer(action="list_peers")

# Terminal 1: Broadcast to ALL DevDucks
🦆 zenoh_peer(action="broadcast", message="git status")
# Both terminals execute and stream responses!

# Send to specific peer
🦆 zenoh_peer(action="send", peer_id="hostname-def456", message="what files are here?")

Cross-Network Connections

Connect DevDuck instances across different networks:

# Machine A (office): Listen for remote connections
export ZENOH_LISTEN="tcp/0.0.0.0:7447"
devduck

# Machine B (home): Connect to office
export ZENOH_CONNECT="tcp/office.example.com:7447"
devduck

# Now they can communicate!
🦆 zenoh_peer(action="broadcast", message="sync all repos")

Use Cases

Scenario	Command	Description
Multi-terminal ops	broadcast "git pull && npm install"	Run on all instances
Distributed tasks	broadcast "analyze ./src"	Parallel analysis
Peer monitoring	list_peers	See all active DevDucks
Direct messaging	send peer_id="..." message="..."	Task specific instance
Cross-network	Set ZENOH_CONNECT	Connect home ↔ office

Environment Variables

Variable	Default	Description
DEVDUCK_ENABLE_ZENOH	true	Auto-start Zenoh on launch
ZENOH_CONNECT	-	Remote endpoint(s) to connect to
ZENOH_LISTEN	-	Endpoint(s) to listen on for remote connections

---

Unified Mesh (CLI + Browser + Cloud)

Connect ALL agent types — terminal DevDucks, browser tabs, and cloud-deployed agents — into a single unified network with shared context.

Architecture

┌─────────────────────────────────────────────────┐
│                 Unified Mesh                     │
│                                                  │
│  ┌──────────┐  ┌──────────┐  ┌──────────────┐  │
│  │ Terminal  │  │ Browser  │  │  AgentCore   │  │
│  │ DevDuck  │  │  Tab(s)  │  │  (Cloud)     │  │
│  │ (Zenoh)  │  │ (WS)     │  │  (AWS)       │  │
│  └────┬─────┘  └────┬─────┘  └──────┬───────┘  │
│       │              │               │           │
│       └──────────────┼───────────────┘           │
│                      │                           │
│           ┌──────────┴──────────┐                │
│           │  mesh_registry.json │                │
│           │  (file-based, TTL)  │                │
│           └─────────────────────┘                │
│                                                  │
│           Ring Context (shared memory)           │
└─────────────────────────────────────────────────┘

Port Allocation

Port	Service	Description
10000	Mesh Relay	AgentCore proxy + browser gateway
10001	WebSocket	Per-message DevDuck server
10002	TCP	Raw socket server
10003	MCP HTTP	Model Context Protocol
10004	IPC	Reserved for Unix socket gateway

Quick Start

# Terminal: Start DevDuck (mesh auto-starts)
devduck
# 🦆 ✓ AgentCore proxy: ws://localhost:10000
# 🦆 ✓ Zenoh peer: hostname-abc123

# Browser: Open mesh.html → auto-connects to ws://localhost:10000
# Browser agents appear in zenoh peer list!

---

AgentCore Deployment (Deploy CLI)

One-command deployment of DevDuck to Amazon Bedrock AgentCore:

Quick Deploy

# Configure and deploy with defaults
devduck deploy --launch

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

# Configure only (don't launch yet)
devduck deploy --name my-agent

# Launch separately
agentcore launch -a my_agent --auto-update-on-conflict

Deploy Options

devduck deploy [OPTIONS]

Options:
  --name, -n          Agent name (default: devduck)
  --tools, -t         Tool config (e.g., "strands_tools:shell,editor")
  --model, -m         Model ID override
  --region, -r        AWS region (default: us-west-2)
  --launch            Auto-launch after configure
  --system-prompt, -s Custom system prompt
  --idle-timeout      Idle timeout seconds (default: 900)
  --max-lifetime      Max lifetime seconds (default: 28800)
  --no-memory         Disable AgentCore memory (STM)
  --no-otel           Disable OpenTelemetry
  --env KEY=VALUE     Additional env vars (repeatable)
  --force-rebuild     Force rebuild dependencies

Manage Deployed Agents

# List all deployed agents
devduck list

# Check agent status
devduck status --name my-agent

# Invoke a deployed agent
devduck invoke "analyze this code" --name my-agent

# Invoke by agent ID directly
devduck invoke "hello" --agent-id abc123-def456

---

Messaging Integrations (Telegram, WhatsApp, Slack)

DevDuck can listen and auto-reply on Telegram, WhatsApp, and Slack — each incoming message spawns a fresh DevDuck instance with full tool access.

Telegram

export TELEGRAM_BOT_TOKEN="your-bot-token"
export STRANDS_TELEGRAM_AUTO_REPLY=true
# Optional: restrict to specific users
export TELEGRAM_ALLOWED_USERS="149632499,cagataycali"

# Start listening for messages
telegram(action="start_listener")

# Send messages
telegram(action="send_message", chat_id="123456", text="Hello!")
telegram(action="send_photo", chat_id="123456", file_path="/path/to/image.png")
telegram(action="send_poll", chat_id="123456", question="Tabs or spaces?", options=["Tabs", "Spaces"])

WhatsApp

Uses wacli — no Cloud API needed, runs via WhatsApp Web protocol.

brew install steipete/tap/wacli && wacli auth
export STRANDS_WHATSAPP_AUTO_REPLY=true

whatsapp(action="start_listener")
whatsapp(action="send_text", to="[phone]", text="Hello from DevDuck!")

Slack

export SLACK_BOT_TOKEN="xoxb-..."
export SLACK_APP_TOKEN="xapp-..."
export STRANDS_SLACK_AUTO_REPLY=true

slack(action="start_listener")
slack(action="send_message", channel="C123", text="Hello!")

---

macOS Integration (use_mac + Apple Notes)

One tool to control your entire Mac — Calendar, Reminders, Mail, Contacts, Safari, Finder, System Events, Shortcuts, Messages, Music, Keychain, and raw AppleScript/JXA.

# Calendar
use_mac(action="calendar.events", days=7)
use_mac(action="calendar.create", title="Meeting", start="2026-03-01 10:00", end="2026-03-01 11:00")

# Reminders
use_mac(action="reminders.create", title="Buy groceries", due_date="2026-03-02", priority=1)

# Mail
use_mac(action="mail.send", to="[e-mail]", subject="Hello", body="Hi there!")
use_mac(action="mail.unread")

# Safari
use_mac(action="safari.tabs")
use_mac(action="safari.open", url="https://strandsagents.com")
use_mac(action="safari.read")  # Read current page text

# System
use_mac(action="system.notify", text="Task complete!", title="DevDuck")
use_mac(action="system.say", text="Hello world", voice="Samantha")
use_mac(action="system.screenshot", path="/tmp/screenshot.png")
use_mac(action="system.volume", level=50)
use_mac(action="system.dark_mode", enable=True)

# Finder
use_mac(action="finder.selection")
use_mac(action="finder.tag", path="/path/to/file", tags="important,review")

# Keychain
use_mac(action="keychain.get", service="MyApp", account="[e-mail]")
use_mac(action="keychain.set", service="MyApp", account="[e-mail]", password="secret")

# Shortcuts
use_mac(action="shortcuts.run", name="My Shortcut", input_text="hello")

# Raw AppleScript / JXA
use_mac(action="applescript", script='tell app "Finder" to get name of every disk')
use_mac(action="jxa", script='Application("System Events").currentUser().name()')

Apple Notes

apple_notes(action="list")
apple_notes(action="create", title="Meeting Notes", body="## Agenda\n- Review Q1")
apple_notes(action="search", query="meeting")
apple_notes(action="export", output_dir="/tmp/my_notes")

---

Spotify Control

Full Spotify playback, search, playlists, queue, library, and discovery control.

export SPOTIFY_CLIENT_ID="your-client-id"
export SPOTIFY_CLIENT_SECRET="your-client-secret"

use_spotify(action="now_playing")
use_spotify(action="search", query="Bohemian Rhapsody", search_type="track")
use_spotify(action="playlists")
use_spotify(action="recommendations", seed_genres="electronic,ambient", limit=10)

---

Reinforcement Learning & ML Toolkit

Train RL agents, run hyperparameter sweeps, fine-tune LLMs — all from DevDuck.

# Train an RL agent
rl(action="train", env_id="CartPole-v1", algorithm="PPO", total_timesteps=50000)

# Evaluate
rl(action="eval", env_id="CartPole-v1", model_path="rl_models/.../best_model")

# Hyperparameter sweep
rl(action="sweep", env_id="LunarLander-v3", n_trials=8)

# Fine-tune LLM with LoRA
rl(action="finetune", model_id="Qwen/Qwen2.5-0.5B", dataset_id="tatsu-lab/alpaca", method="lora")

# List saved models
rl(action="list_models")

---

Advanced Features

🎬 Session Recording (Time-Travel Debugging)

Record entire sessions for replay, debugging, and state restoration:

# CLI: Start with recording enabled
devduck --record
devduck --record "analyze this codebase"

# Resume from recorded session
devduck --resume ~/Desktop/session-20250202-123456.zip
devduck --resume session.zip "continue where we left off"
devduck --resume session.zip --snapshot 2 "what was I working on?"

Interactive recording:

🦆 record              # Toggle recording on/off
🦆 session_recorder(action="start")
🦆 session_recorder(action="snapshot", description="before refactor")
🦆 session_recorder(action="stop")  # Exports to /tmp/devduck/recordings/

Captures three layers:

• sys: OS-level events (file I/O, HTTP requests)
• tool: All tool calls and results
• agent: Messages, decisions, state changes

Python API for session analysis:

from devduck import load_session, resume_session, list_sessions

# Load and analyze a session
session = load_session("~/Desktop/session-20250202-123456.zip")
print(session)  # LoadedSession(events=156, snapshots=3, duration=342.5s)

# Resume from snapshot (restores conversation history!)
result = session.resume_from_snapshot(2, agent=devduck.agent)
print(f"Restored {result['messages_restored']} messages")

# Resume and continue with new query
result = session.resume_and_continue(2, "what files did we modify?", devduck.agent)

Recordings saved to: /tmp/devduck/recordings/

---

🎥 Asciinema Recording (Shareable Agent Trajectories)

Record agent sessions as .cast files — play them anywhere with asciinema play, upload to asciinema.org, or embed in docs.

# Enable asciinema recording
DEVDUCK_ASCIINEMA=true devduck

# Custom output directory
DEVDUCK_CAST_DIR=~/my-casts DEVDUCK_ASCIINEMA=true devduck

What gets captured:

Event	Description
🦆 User prompts	Your questions recorded as input events
📝 Agent streaming	Token-by-token text output
🛠️ Tool markers	Visual ─── separators when tools start
✔/✖ Tool results	Completion/failure with durations
💭 Thinking	Reasoning/thinking text (dimmed)

Playback:

asciinema play /tmp/devduck/casts/devduck-20260306-023100.cast
asciinema upload /tmp/devduck/casts/devduck-20260306-023100.cast

---

🌙 Ambient Mode (Background Thinking)

Continue working in the background while you're idle:

# Enable via environment
export DEVDUCK_AMBIENT_MODE=true
devduck

# Or toggle in REPL
🦆 ambient     # Toggle standard ambient mode
🦆 auto        # Toggle autonomous mode

Standard Mode: Runs up to 3 iterations when you go idle (30s)

Autonomous Mode: Runs continuously until done or stopped — agent signals completion with [AMBIENT_DONE]

How it works:

1. You go idle (30s default)
2. DevDuck continues exploring the last topic
3. Background work streams with 🌙 prefix
4. When you return, findings are injected into your next query

---

System Prompt Management

Self-improvement - agent updates its own system prompt:

# View current system prompt
system_prompt(action="view")

# Add new context (appends to prompt)
system_prompt(action="add_context", context="New learning: Always use FastAPI for APIs")

# Update entire prompt
system_prompt(action="update", prompt="You are a specialized DevOps agent...")

# Sync to GitHub (persist across deployments)
system_prompt(action="update", prompt="Updated...", repository="cagataycali/devduck")

---

Access Methods

Protocol	Endpoint	Test Command	Use Case
CLI	Terminal	devduck "query"	Interactive/one-shot
TUI	Terminal	devduck --tui	Multi-conversation UI
Python	Import	import devduck; devduck("query")	Script integration
Mesh Relay	localhost:10000	Open mesh.html	Unified mesh (browser + all agents)
WebSocket	localhost:10001	wscat -c ws://localhost:10001	Browser/async apps
TCP	localhost:10002	nc localhost 10002	Network clients
MCP	localhost:10003/mcp	Add to Claude Desktop	MCP clients
IPC	/tmp/devduck_main.sock	nc -U /tmp/devduck_main.sock	Local processes

CLI Commands

devduck                                   # Interactive REPL
devduck --tui                             # Multi-conversation TUI
devduck "your query here"                 # One-shot query
devduck --mcp                             # MCP stdio mode (Claude Desktop)
devduck --record                          # Start with recording enabled
devduck --resume session.zip              # Resume from recorded session
devduck deploy --launch                   # Deploy to AgentCore
devduck list                              # List deployed agents
devduck status --name my-agent            # Check agent status
devduck invoke "hello" --name my-agent    # Invoke deployed agent

REPL Commands

Command	Description
exit / quit / q	Exit DevDuck
ambient	Toggle standard ambient mode
auto / autonomous	Toggle autonomous mode
record	Toggle session recording
!<command>	Execute shell command (e.g., !ls -la)

---

Configuration

Variable	Default	Description
Model
MODEL_PROVIDER	Auto	Manual override: bedrock, anthropic, openai, github, gemini, cohere, writer, mistral, litellm, llamaapi, mlx, ollama
STRANDS_MODEL_ID	Auto	Model name (e.g., claude-sonnet-4, gpt-4o, qwen3:1.7b)
Provider API Keys
ANTHROPIC_API_KEY	-	Anthropic API key (auto-detected)
OPENAI_API_KEY	-	OpenAI API key (auto-detected)
GOOGLE_API_KEY / GEMINI_API_KEY	-	Google Gemini API key (auto-detected)
GITHUB_TOKEN / PAT_TOKEN	-	GitHub token for GitHub Models (auto-detected)
COHERE_API_KEY	-	Cohere API key (auto-detected)
WRITER_API_KEY	-	Writer API key (auto-detected)
MISTRAL_API_KEY	-	Mistral API key (auto-detected)
LITELLM_API_KEY	-	LiteLLM API key (auto-detected)
LLAMAAPI_API_KEY	-	LlamaAPI key (auto-detected)
Tools
DEVDUCK_TOOLS	60+ tools	Format: package1:tool1,tool2;package2:tool3
DEVDUCK_LOAD_TOOLS_FROM_DIR	false	Auto-load from ./tools/ directory
Memory
DEVDUCK_KNOWLEDGE_BASE_ID	-	Bedrock KB ID for auto-RAG
SYSTEM_PROMPT	-	Additional system prompt content
MCP
MCP_SERVERS	-	JSON config for external MCP servers
Servers
DEVDUCK_TCP_PORT	10002	TCP server port
DEVDUCK_WS_PORT	10001	WebSocket server port
DEVDUCK_MCP_PORT	10003	MCP server port
DEVDUCK_AGENTCORE_PROXY_PORT	10000	Mesh relay / AgentCore proxy port
DEVDUCK_IPC_SOCKET	/tmp/devduck_main.sock	IPC socket path
DEVDUCK_ENABLE_TCP	false	Enable TCP server
DEVDUCK_ENABLE_WS	true	Enable WebSocket server
DEVDUCK_ENABLE_MCP	false	Enable MCP server
DEVDUCK_ENABLE_IPC	false	Enable IPC server
DEVDUCK_ENABLE_ZENOH	true	Enable Zenoh peer-to-peer
DEVDUCK_ENABLE_AGENTCORE_PROXY	true	Enable unified mesh relay
ZENOH_CONNECT	-	Remote Zenoh endpoint(s) to connect to
ZENOH_LISTEN	-	Zenoh endpoint(s) to listen on
Ambient Mode
DEVDUCK_AMBIENT_MODE	false	Enable ambient mode on startup
DEVDUCK_AMBIENT_IDLE_SECONDS	30	Seconds idle before ambient starts
DEVDUCK_AMBIENT_MAX_ITERATIONS	3	Max iterations in standard mode
DEVDUCK_AMBIENT_COOLDOWN	60	Seconds between ambient runs
DEVDUCK_AUTONOMOUS_MAX_ITERATIONS	100	Max iterations in autonomous mode
DEVDUCK_AUTONOMOUS_COOLDOWN	10	Seconds between autonomous runs
LSP
DEVDUCK_LSP_AUTO_DIAGNOSTICS	false	Auto-append LSP diagnostics after file edits
Recording
DEVDUCK_ASCIINEMA	false	Enable asciinema .cast recording
DEVDUCK_CAST_DIR	/tmp/devduck/casts	Output directory for .cast files
Messaging
TELEGRAM_BOT_TOKEN	-	Telegram bot token from @BotFather
STRANDS_TELEGRAM_AUTO_REPLY	false	Enable auto-reply on Telegram
TELEGRAM_ALLOWED_USERS	-	Comma-separated user IDs/usernames allowlist
SLACK_BOT_TOKEN	-	Slack bot token (xoxb-...)
SLACK_APP_TOKEN	-	Slack app token for Socket Mode (xapp-...)
STRANDS_SLACK_AUTO_REPLY	false	Enable auto-reply on Slack
STRANDS_WHATSAPP_AUTO_REPLY	false	Enable auto-reply on WhatsApp
Spotify
SPOTIFY_CLIENT_ID	-	Spotify client ID
SPOTIFY_CLIENT_SECRET	-	Spotify client secret
Context
DEVDUCK_LOG_LINE_COUNT	50	Recent log lines in context
DEVDUCK_LAST_MESSAGE_COUNT	200	Recent messages in context

---

Troubleshooting

Ollama model not found:

ollama pull qwen3:1.7b

Port already in use:

export DEVDUCK_AGENTCORE_PROXY_PORT=10010
export DEVDUCK_WS_PORT=10011
devduck

Context window overflow: DevDuck auto-recovers by clearing history and retrying. If it persists:

manage_messages(action="clear")

Hot-reload not working:

mkdir -p ./tools
devduck
🦆 view_logs(action="search", pattern="watcher")

Memory/performance issues:

export STRANDS_MODEL_ID="qwen3:0.5b"
export DEVDUCK_LOG_LINE_COUNT=20
export DEVDUCK_LAST_MESSAGE_COUNT=50

View logs: devduck → 🦆 view_logs()

---

📋 All Built-in Tools (60+)

DevDuck Core Tools

• system_prompt - Update agent's system prompt (GitHub sync support)
• store_in_kb - Store content in Bedrock Knowledge Base
• state_manager - Save/restore agent state (time-travel)
• session_recorder - 🎬 Record sessions for replay and debugging
• manage_tools - Runtime tool add/remove/create/fetch/discover
• manage_messages - View/drop/compact/export conversation history
• tasks - 📋 Background parallel agent tasks
• scheduler - ⏰ Cron and one-time job scheduling
• sqlite_memory - 🧠 SQLite persistent memory with FTS
• dialog - 💬 Interactive terminal dialogs (forms, pickers)
• notify - 🔔 Native OS notifications
• use_computer - 🖥️ Mouse, keyboard, screenshots
• listen - 🎤 Background speech transcription (Whisper)
• lsp - 🔍 Language Server Protocol integration
• tui - 🖥️ Push content to TUI from agent
• tcp - TCP server with real-time streaming
• websocket - WebSocket server with concurrent messaging
• ipc - Unix socket IPC server for local processes
• mcp_server - Expose as MCP server (HTTP/stdio)
• zenoh_peer - Peer-to-peer networking with auto-discovery
• agentcore_proxy - 🌐 Unified mesh relay (Zenoh + AgentCore + browser peers)
• unified_mesh - Single source of truth for all agent types + ring context
• mesh_registry - File-based agent discovery with fcntl locking and TTL
• ambient_mode - Control ambient/autonomous background thinking
• install_tools - Install packages and load tools at runtime
• create_subagent - Spawn sub-agents via GitHub Actions
• use_github - GitHub GraphQL API operations
• fetch_github_tool - Fetch and load tools from GitHub repos
• gist - GitHub Gist management (create/update/fork/star/comment)
• tray - System tray app control (macOS)
• ambient - Ambient AI input overlay (macOS)
• agentcore_config - Configure & launch on Bedrock AgentCore
• agentcore_invoke - Invoke deployed AgentCore agents
• agentcore_logs - View CloudWatch logs from agents
• agentcore_agents - List/manage agent runtimes
• view_logs - View/search/clear DevDuck logs
• speech_to_speech - Real-time speech-to-speech conversations
• use_mac - 🍎 Unified macOS system control
• apple_notes - 📝 Apple Notes management
• use_spotify - 🎵 Full Spotify control
• telegram - 💬 Telegram bot integration
• whatsapp - 💬 WhatsApp integration via wacli
• slack - 💬 Slack integration
• jsonrpc - 🔌 Generic JSON-RPC client
• rl - 🎮 RL & ML toolkit
• scraper - HTML/XML parsing with BeautifulSoup4

Apple Silicon Tools (./tools/)

• apple_nlp - 🧠 On-device NLP (sentiment, entities, embeddings, POS)
• apple_vision - 👁️ On-device Vision (OCR, barcodes, faces, rectangles)
• apple_wifi - 📡 WiFi intelligence (scan, signal, diagnostics)
• apple_sensors - 🍎 Hardware sensors (accelerometer, ambient light, lid)
• apple_smc - 🌡️ SMC thermal, fan, and power data

Strands Tools

• shell - Interactive shell with PTY support
• editor - File editing (view/create/replace/insert/undo)
• file_read - Multi-file reading with search modes
• file_write - Write content to files
• calculator - SymPy-powered math
• image_reader - Read images for Converse API
• use_agent - Nested agent with different model
• retrieve - Bedrock Knowledge Base retrieval
• speak - Text-to-speech

Community / Hot-Reload Tools

• add_comment - Add comments to GitHub issues/PRs
• list_issues - List GitHub repository issues
• list_pull_requests - List GitHub repository PRs
• Hot-reload any .py from ./tools/ when DEVDUCK_LOAD_TOOLS_FROM_DIR=true

---

GitHub Actions

Run DevDuck in CI/CD pipelines:

name: AI Code Assistant
on: 
  issues:
    types: [opened, edited]
  pull_request:
    types: [opened, edited, synchronize]

jobs:
  devduck:
    runs-on: ubuntu-latest
    permissions:
      contents: read
      issues: write
      pull-requests: write
    steps:
      - uses: cagataycali/devduck@main
        with:
          task: "Analyze and help with this issue or PR"
          provider: "github"
          model: "gpt-4o"
          tools: "shell,file_read,file_write,use_github,calculator"
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

---

Resources

• Strands SDK: github.com/strands-agents/sdk-python
• Documentation: strandsagents.com
• Web UI: cagataycali.github.io/devduck

---

Citation

@software{devduck2025,
  author = {Cagatay Cali},
  title = {DevDuck: Self-Modifying AI Agent with Unified Mesh, Hot-Reload, and Multi-Protocol Servers},
  year = {2025},
  url = {https://github.com/cagataycali/devduck}
}

---

Apache 2.0 | Built with Strands Agents | @cagataycali