Repolode
Submit
razzant

razzant/ouroboros

Open-source self-modifying AI agent that keeps its identity and memory across restarts, works on external projects, and can rewrite its own code.

View on GitHubWebsite
Stars
1,612
Forks
689
Open issues
216
Last commit
18 hours ago

What is it?

Ouroboros is a general-purpose AI agent whose identity, durable memory and history persist across tasks and restarts. It can work on external projects, coordinate a set of specialist subagents in parallel, and modify its own implementation, including code, prompts, tools and dependencies. Changes to its own body go through deterministic checks and independent review, and remain part of its Git history.

It runs as a native desktop app for macOS, Windows and Linux, or as a headless CLI and web server. Its repository, memory and interface stay on the user's machine. Model inference can use remote provider APIs or a local GGUF model, and delegated coding can use connected Codex, Claude Code or Cursor accounts through its bundled Claudexor engine.

The project also supports background reflection between requests, scheduled tasks, reviewed extensions such as skills and MCP tools, and Docker deployment for the web runtime. It is MIT licensed and written primarily in Python.

Key features

  • 01Keeps identity, memory and history across tasks and restarts.
  • 02Modifies its own code, prompts, tools and dependencies, with reviewed changes tracked in Git.
  • 03Coordinates parallel specialist subagents and shares findings across a task tree.
  • 04Works on external project folders or Git worktrees and exports changes as patch files.
  • 05Runs as a desktop app, a CLI, or a Docker web server, with remote or local model options.

Use cases

  • Delegating multi-step coding or investigation tasks to an agent that retains project context.
  • Running a self-hosted, always-on agent with persistent memory on a personal machine.
  • Scheduling recurring maintenance or review tasks through the CLI.
  • Experimenting with agents that evolve their own codebase under review controls.

Quick start

  1. 1

    Download the desktop app for your platform from the releases page or the download links above

  2. 2

    On macOS, open the DMG and drag Ouroboros.app to Applications, then open it

  3. 3

    On Debian or Ubuntu, install the downloaded package

    sudo apt install ./ouroboros_*_amd64.deb
  4. 4

    On Windows, extract the ZIP and run Ouroboros.exe

Things to know

  • Desktop downloads are only provided for macOS on Apple silicon, Windows x64, and Linux x86_64.
  • Running tasks requires a Codex account, a supported remote provider API key, or a local GGUF model.
  • Android support is experimental and limited to Magisk-rooted ARM64 devices.
  • Benchmark results are self-reported by the project.

In collections

AI coding agentsOpen-source AI agent frameworks & appsAI coding assistants & agentsMCP serversSelf-hosted apps

Topics

#ai-agent#llm#coding-agent#self-hosted#desktop-agent#mcp#persistent-memory#python#agentic-ai#ai-agents

FAQ

What is ouroboros?
Open-source self-modifying AI agent that keeps its identity and memory across restarts, works on external projects, and can rewrite its own code. Ouroboros is a general-purpose AI agent whose identity, durable memory and history persist across tasks and restarts. It can work on external projects, coordinate a set of specialist subagents in parallel, and modify its own implementation, including code, prompts, tools and dependencies. Changes to its own body go through deterministic checks and independent review, and remain part of its Git history.
Is ouroboros free to use? What license does it have?
ouroboros is open source under the MIT license. Check the license terms before using it in commercial products.
Is ouroboros still maintained?
Actively maintained. The last commit to ouroboros was on Nov 6, 2026.
What can I use ouroboros for?
Delegating multi-step coding or investigation tasks to an agent that retains project context. Running a self-hosted, always-on agent with persistent memory on a personal machine. Scheduling recurring maintenance or review tasks through the CLI. Experimenting with agents that evolve their own codebase under review controls.
How do I install ouroboros?
Download the desktop app for your platform from the releases page or the download links above On macOS, open the DMG and drag Ouroboros.app to Applications, then open it On Debian or Ubuntu, install the downloaded package: sudo apt install ./ouroboros_*_amd64.deb On Windows, extract the ZIP and run Ouroboros.exe
Where can I get ouroboros?
The source code and installation instructions are on GitHub: https://github.com/razzant/ouroboros

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

Related projects