Extreme minimalist self-adapting AI agent
A single-file AI agent that's self-aware, self-healing, and ridiculously powerful.
Record complete sessions with three-layer capture (sys/tool/agent). Resume from any snapshot with full conversation state restored.
Learn more →Background thinking while you're idle. Standard mode explores topics; autonomous mode builds entire features.
Learn more →Modify the agent while it's running. Code changes apply instantly without restart. Protected during execution.
Multiple DevDuck instances auto-discover each other. Broadcast commands to all or send to specific peers.
Learn more →Load tools from any Python package at runtime. Fetch tools from GitHub URLs. MCP server integration.
Learn more →Full access to its own source code. Can read, understand, and modify itself. Answers questions about its implementation.
Automatic retrieval and storage with Amazon Bedrock Knowledge Bases. Seamless memory across sessions.
Learn more →Built-in WebSocket server for real-time streaming. Connect from browsers, apps, or other agents.
Expose DevDuck as MCP server for Claude Desktop, or load external MCP servers to extend capabilities.
Learn more →One-click setup for Linux, macOS, Windows, Docker, Android, and Cloud. Generate custom install scripts.
Setup wizard →Modify the agent while it's running. Code changes apply instantly without restart. Protected during execution.
Learn more →Get DevDuck running in under a minute.
pipx install devduckdevduckdevduck "build me a web scraper"# Python API
from devduck import ask, status, devduck
# Quick query
result = ask("create a REST API")
# Direct agent access
result = devduck("analyze this codebase")
# Session recording
from devduck import start_recording, stop_recording, load_session
start_recording()
# ... work ...
export_path = stop_recording()
# Resume session
session = load_session("session.zip")
session.resume_from_snapshot(2, agent=devduck.agent)Connect to a running DevDuck instance via WebSocket.
Time-travel debugging for AI agents with full state restoration.
Capture complete agent state including conversation history, system prompt, model info, and working directory.
Full agent.messages array stored in snapshots. Resume with complete context—like hitting pause and play.
Sys (file I/O, network), Tool (calls & results), Agent (messages, decisions). Complete visibility.