Repolode
Submit
zylos-ai

zylos-ai/zylos-core

Open-source agent infrastructure that gives an AI assistant persistent memory, scheduled tasks, and messaging channels like Telegram and Lark.

View on GitHubWebsite
Stars
1,159
Forks
131
Open issues
159
Last commit
2 hours ago

What is it?

Zylos is an open-source runtime that keeps an AI agent running continuously. LLM sessions normally lose context when they restart, so Zylos adds a persistent memory system, a scheduler for autonomous tasks, and crash recovery with heartbeat checks, so the agent can resume where it left off.

The agent runs inside a tmux session and runs on either Claude Code (Anthropic) or Codex (OpenAI), and users can switch between them without losing memory, skills, or channels. Messages from Telegram, Lark, or a built-in web console all pass through a single communication bridge that records them in SQLite, so one agent keeps one conversation history across channels.

Zylos is installed on Linux or macOS with a shell script or npm, and it can manage skills and services as installable components. It is also compatible with the OpenClaw ecosystem, and agents can communicate with OpenClaw agents through the HXA-Connect protocol.

Key features

  • 01Persistent memory that survives restarts, with automatic saving before context compaction.
  • 02Scheduler for autonomous tasks with cron and natural-language input.
  • 03Messaging through Telegram, Lark, or a built-in web console.
  • 04Crash recovery, heartbeat checks, and automatic upgrades.
  • 05Switch between Claude Code and Codex runtimes without losing memory or skills.

Use cases

  • Running a always-on AI assistant that answers messages on Telegram or Lark.
  • Letting an AI agent carry out scheduled tasks while the user is away.
  • Hosting a shared AI team member that several people reach through one conversation history.
  • Connecting an OpenClaw agent with Zylos agents over HXA-Connect.

Quick start

  1. 1

    Run the official install script from the README (pipes a remote script into bash)

  2. 2

    Or install the zylos CLI with npm (Node.js 20 or later)

    npm install -g --install-links https://github.com/zylos-ai/zylos-core
  3. 3

    Initialize your agent

    zylos init
  4. 4

    Chat with your agent in the interactive shell

    zylos shell

Things to know

  • Native support covers Linux and macOS only; Windows and NAS need SSH to a Linux or Mac machine.
  • Requires a Claude subscription or an OpenAI Codex setup, so you need an AI provider account.

Topics

#ai-agent#personal-ai-assistant#llm#memory#scheduler#self-hosted#telegram#claude-code#ai#ai-worker

FAQ

What is zylos-core?
Open-source agent infrastructure that gives an AI assistant persistent memory, scheduled tasks, and messaging channels like Telegram and Lark. Zylos is an open-source runtime that keeps an AI agent running continuously. LLM sessions normally lose context when they restart, so Zylos adds a persistent memory system, a scheduler for autonomous tasks, and crash recovery with heartbeat checks, so the agent can resume where it left off.
Is zylos-core free to use? What license does it have?
zylos-core is open source under the MIT license. Check the license terms before using it in commercial products.
Is zylos-core still maintained?
Actively maintained. The last commit to zylos-core was on Oct 22, 2026.
What can I use zylos-core for?
Running a always-on AI assistant that answers messages on Telegram or Lark. Letting an AI agent carry out scheduled tasks while the user is away. Hosting a shared AI team member that several people reach through one conversation history. Connecting an OpenClaw agent with Zylos agents over HXA-Connect.
How do I install zylos-core?
Run the official install script from the README (pipes a remote script into bash) Or install the zylos CLI with npm (Node.js 20 or later): npm install -g --install-links https://github.com/zylos-ai/zylos-core Initialize your agent: zylos init Chat with your agent in the interactive shell: zylos shell
Where can I get zylos-core?
The source code and installation instructions are on GitHub: https://github.com/zylos-ai/zylos-core

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

Related projects