Q00/ouroboros

Agent OS: the agent gets smarter on its own. We just hold the line: Interview-gated, staged evaluation, budgeted evolution loop. MCP server, 14 runtimes: Claude Code, Codex CLI, Gemini CLI, OpenCode, Copilot, Kiro and more.

What it solves

Ouroboros addresses the problem of non-deterministic AI coding where vague prompts lead to AI guessing, architectural drift, and manual QA bottlenecks. It shifts the focus from prompt engineering to a structured, specification-first workflow that ensures human clarity before code is written.

How it works

Ouroboros acts as an "Agent OS" or local-first runtime layer that sits between the user and various AI coding agents (such as Claude Code, Codex CLI, and GitHub Copilot CLI). It transforms vague requests into a verified codebase through a five-stage cycle:

  1. Interview: A Socratic interview process that exposes hidden assumptions and reduces ambiguity.
  2. Crystallize: Locking the intent into an immutable "Seed" specification.
  3. Execute: Running the agent to build the codebase based on the locked spec.
  4. Evaluate: A three-stage automated evaluation gate to verify the output.
  5. Evolve: Refining the project based on evaluation results.

Who it’s for

Developers using AI coding agents who want to move away from ad-hoc prompting and toward replayable, observable, and policy-bound execution contracts.

Highlights

  • Multi-Runtime Support: Works across a wide array of agents including Claude Code, Codex, Gemini, Copilot, DeepSeek, and more.
  • Agent OS Architecture: Split into a shell (Ourocode), an application layer (plugins), and a core kernel (this repository) for managing Seeds, Ledgers, and Runtimes.
  • Specification-First: Replaces guessing with a Socratic interview and an immutable seed spec to prevent architectural drift.
  • MCP Integration: Registers as an MCP server to integrate directly into supported AI coding hosts.

Related

  • Project
    razzant/ouroboros

    An open-source, general-purpose AI agent with durable memory and a self-evolving core that can rewrite its own code and architecture to improve itself.

  • Project
    EverMind-AI/EverOS

    EverOS is a Python library that provides local‑first, Markdown‑backed memory for AI agents. It syncs files to SQLite and LanceDB, offers a simple HTTP API for add/flush/search, and works with a single OpenRouter key for keyword search. Optional extras add embedding‑based hybrid search, reranking, and multimodal file ingestion.

  • Project
    rlaope/oh-my-hermes

    An operating layer for Hermes Agent that adds structured workflow management, intelligent model routing, parallel task execution, and a persistent memory system to improve agent reliability and efficiency.

  • Project
    jonwiggins/optio

    A self-hosted AI agent orchestration platform that runs agent swarms and workflows on your own Kubernetes cluster or local machines.

  • Project
    Ancienttwo/repo-harness

    A file-backed workflow and runtime for AI agents that replaces chat memory with repository-based state, enabling resumable sessions and token-efficient development.