Repolode
Submit
Vaan21th

Vaan21th/daemonkey-agent

A local-first AI companion daemon for Windows and macOS that keeps long-term memory of you, turns chats into office files, and can extend itself.

View on GitHub
Stars
94
Forks
13
Open issues
3
Last commit
16 days ago

What is it?

Daemonkey runs as a background program on your own computer and connects to any OpenAI-compatible model API. Its memory is split into two stores: a soul layer with its identity, your profile and a growth diary, and a playbook layer that saves procedures from tasks it has completed so similar work next time starts with that know-how. Both are plain files on your disk that you can read, edit and move.

The interface includes a workbench where you chat on the left and open PPT, Word or Excel files in the middle, marking up a selected passage or pinning a note so only that part changes. A desktop pet and a room view share the same memory, and WeChat and Feishu can deliver messages in. Scheduled tasks run at set times, and a repair console and rollback script help recover after a broken change.

The project documents prompt-cache-friendly design: a stable system prefix with volatile content placed at the end, on-demand tool loading, and a layered profile. It also reports a memory benchmark result on ATM-Bench, with the methodology described in the project's own materials.

Key features

  • 01Long-term memory stored locally as readable files, split into profile and playbook layers.
  • 02Workbench for opening and marking up PPT, Word and Excel documents in chat.
  • 03Desktop pet and room view sharing one memory.
  • 04WeChat and Feishu message integration and scheduled tasks.
  • 05Works with any OpenAI-compatible API, including proxies.
  • 06Repair console and one-click rollback to a working kernel version.

Use cases

  • A personal assistant that keeps context about your work and preferences across model changes.
  • Drafting and revising slide decks or documents by marking up only the parts you select.
  • Running daily reports or summaries on a schedule while you are away from the computer.
  • Experimenting with self-hosted agent memory on a local machine.

Quick start

  1. 1

    Install Python 3.10+ (tick Add to PATH)

  2. 2

    Windows: double-click Daemonkey.exe, then Environment, then Install

  3. 3

    macOS / Linux: run the start script from the project folder

    chmod +x start.sh && ./start.sh
  4. 4

    Start it, then enter an OpenAI-compatible API key in the browser

Things to know

  • Officially supports only Windows and macOS.
  • Needs an OpenAI-compatible API key, which may cost money.
  • Licensed under AGPL-3.0, so modified versions offered over a network must be open-sourced.

Topics

#personal-agent#ai-companion#agent-memory#local-first#long-term-memory#llm#self-hosted#python#agent-harness#agentic

FAQ

What is daemonkey-agent?
A local-first AI companion daemon for Windows and macOS that keeps long-term memory of you, turns chats into office files, and can extend itself. Daemonkey runs as a background program on your own computer and connects to any OpenAI-compatible model API. Its memory is split into two stores: a soul layer with its identity, your profile and a growth diary, and a playbook layer that saves procedures from tasks it has completed so similar work next time starts with that know-how. Both are plain files on your disk that you can read, edit and move.
Is daemonkey-agent free to use? What license does it have?
daemonkey-agent is open source under the AGPL-3.0 license. Check the license terms before using it in commercial products.
Is daemonkey-agent still maintained?
Actively maintained. The last commit to daemonkey-agent was on Oct 6, 2026.
What can I use daemonkey-agent for?
A personal assistant that keeps context about your work and preferences across model changes. Drafting and revising slide decks or documents by marking up only the parts you select. Running daily reports or summaries on a schedule while you are away from the computer. Experimenting with self-hosted agent memory on a local machine.
How do I install daemonkey-agent?
Install Python 3.10+ (tick Add to PATH) Windows: double-click Daemonkey.exe, then Environment, then Install macOS / Linux: run the start script from the project folder: chmod +x start.sh && ./start.sh Start it, then enter an OpenAI-compatible API key in the browser
Where can I get daemonkey-agent?
The source code and installation instructions are on GitHub: https://github.com/Vaan21th/daemonkey-agent

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects