Cognitive Os

Cognitive OS — AI Agent Operating System & Multi-Harness Specification

[](https://github.com/aegntic/cognitive-os/actions/workflows/ci.yml) [](https://github.com/aegntic/cognitive-os/actions/workflows/codeql.yml) [](LICENSE) [](https://github.com/aegntic/cognitive-os/actions/workflows/ci.yml)

**The future is not predicted. It is compiled.**

**Cognitive OS** is a universal cognitive operating system specification for AI coding agents. It defines a production-grade, 16-section cognitive architecture that governs how LLM-powered agents think, plan, verify, and execute across real-world systems. Unlike single-harness prompt files, Cognitive OS auto-adapts to **Claude Code**, **Cursor**, **Codex CLI**, **Gemini CLI**, and **Factory Droid** through a single master specification.

---

Table of Contents

1. Why Cognitive OS?
2. Key Features
3. What Problem Does This Solve?
4. Harness Adapters
5. Quick Start
6. Architecture Overview
7. Build & Test Commands
8. Comparison: Cognitive OS vs Other Approaches
9. FAQ
10. File Structure
11. Contributing
12. Changelog
13. Security
14. License

---

Why Cognitive OS?

AI coding agents (Claude Code, Cursor, Codex, Gemini, Factory Droid) share the same fundamental problems: context window exhaustion, unverified outputs, prompt injection vulnerabilities, inconsistent reasoning quality, and untracked token costs. Existing solutions are fragm