API Reference¶

Python API¶

Module-Level Functions¶

import devduck

# Direct call (module is callable)
devduck("your query")

# Helper functions
from devduck import ask, status, restart, hot_reload

result = ask("analyze this code")
info = status()
restart()
hot_reload()

Session Recording API¶

from devduck import (
    start_recording,
    stop_recording,
    load_session,
    resume_session,
    list_sessions,
)

# Record
recorder = start_recording(session_id="my-session")
export_path = stop_recording()

# Load and replay
session = load_session("session.zip")
print(session.events)
print(session.snapshots)

# Resume
result = resume_session("session.zip", new_query="continue")
result = resume_session("session.zip", snapshot_id=2)

# List
sessions = list_sessions()

DevDuck Class¶

from devduck import devduck

# The auto-initialized instance
devduck("query")
devduck.status()
devduck.restart()
devduck.agent  # Strands Agent instance
devduck.model  # Current model name
devduck.tools  # Loaded tools list
devduck.ambient  # AmbientMode instance (if enabled)

CLI Reference¶

devduck [query]              # Interactive REPL or one-shot
devduck --tui                # Multi-conversation TUI
devduck --mcp                # MCP stdio mode
devduck --record             # Enable session recording
devduck --resume FILE        # Resume from session
devduck --snapshot ID        # Specific snapshot (with --resume)

# AgentCore
devduck deploy [options]     # Deploy to AgentCore
devduck list [--region]      # List deployed agents
devduck status [--name]      # Check agent status
devduck invoke PROMPT [opts] # Invoke deployed agent

Tool Quick Reference¶

Core Tools¶

Tool	Key Actions
`shell(command)`	Execute shell commands
`editor(command, path, ...)`	File editing
`file_read(path, mode)`	Read files
`file_write(path, content)`	Write files
`use_github(query, variables)`	GitHub GraphQL API
`use_agent(prompt, system_prompt, ...)`	Spawn sub-agents
`manage_tools(action, ...)`	Runtime tool management
`manage_messages(action, ...)`	Conversation history

DevDuck Tools¶

Tool	Key Actions
`system_prompt(action, ...)`	View/update system prompt
`websocket(action, ...)`	WebSocket server
`zenoh_peer(action, ...)`	P2P networking
`agentcore_proxy(action, ...)`	Mesh relay
`tasks(action, prompt, ...)`	Background tasks
`scheduler(action, name, ...)`	Cron jobs
`session_recorder(action, ...)`	Session recording
`ambient_mode(action, ...)`	Background thinking
`sqlite_memory(action, ...)`	Persistent memory
`telegram(action, ...)`	Telegram bot
`dialog(dialog_type, text, ...)`	Interactive dialogs
`use_computer(action, ...)`	Mouse/keyboard/screenshots
`listen(action, ...)`	Speech-to-text
`retrieve(text, ...)`	Knowledge base RAG
`store_in_kb(content, ...)`	Knowledge base storage

Environment Variables¶

Model¶

Variable	Description
`MODEL_PROVIDER`	Force model provider
`STRANDS_MODEL_ID`	Specific model ID
`STRANDS_MAX_TOKENS`	Max tokens (default: 60000)
`STRANDS_TEMPERATURE`	Temperature (default: 1.0)
`ANTHROPIC_API_KEY`	Anthropic API key
`OPENAI_API_KEY`	OpenAI API key
`GOOGLE_API_KEY`	Gemini API key
`OLLAMA_HOST`	Ollama host URL

Servers¶

Variable	Default	Description
`DEVDUCK_ENABLE_WS`	`true`	WebSocket server
`DEVDUCK_WS_PORT`	`10001`	WebSocket port
`DEVDUCK_ENABLE_TCP`	`false`	TCP server
`DEVDUCK_TCP_PORT`	`10002`	TCP port
`DEVDUCK_ENABLE_MCP`	`false`	MCP HTTP server
`DEVDUCK_MCP_PORT`	`10003`	MCP port
`DEVDUCK_ENABLE_ZENOH`	`true`	Zenoh P2P
`DEVDUCK_ENABLE_AGENTCORE_PROXY`	`true`	Mesh relay
`DEVDUCK_AGENTCORE_PROXY_PORT`	`10000`	Proxy port

Features¶

Variable	Default	Description
`DEVDUCK_TOOLS`	core set	Tool configuration
`DEVDUCK_AMBIENT_MODE`	`false`	Ambient mode
`DEVDUCK_KNOWLEDGE_BASE_ID`	—	Auto RAG
`DEVDUCK_LOAD_TOOLS_FROM_DIR`	`true`	Hot-load ./tools/
`DEVDUCK_ASCIINEMA`	`false`	Record .cast
`DEVDUCK_AUTO_START_SERVERS`	`true`	Auto-start servers
`MCP_SERVERS`	—	MCP server config (JSON)
`BYPASS_TOOL_CONSENT`	`true`	Skip tool confirmations