Intelligence That Answers to You.
Covenant AI is an open-source artificial intelligence infrastructure platform built around the principle that AI systems should always remain accountable to the humans they serve. Unlike traditional AI frameworks that prioritize unrestricted automation and persistent agent autonomy, Covenant AI is designed with enforceable human authority, bounded execution, transparency, and verifiable decision-making at its core. Every action performed by the system is intended to be authorized, traceable, explainable, and revocable, creating a foundation for trustworthy and human-centered AI infrastructure.
The platform includes a comprehensive authority and intent framework that allows users to define exactly what an AI system is permitted to do, for how long, and under what conditions. Covenant AI uses intent-scoped permissions, temporary delegation tokens, and constraint-driven execution environments to prevent unauthorized escalation or uncontrolled autonomy. Features such as the Intent Receipt Protocol (IRP), human-readable audit trails, cryptographic verification, and explainable execution summaries provide users and organizations with complete visibility into how decisions are made and actions are performed.
Security and privacy are foundational components of the architecture. Covenant AI incorporates zero-trust infrastructure, sandboxed agent runtimes, encrypted delegation systems, and local-first privacy controls to minimize unnecessary data exposure and reduce systemic risk. The platform supports self-hosted deployments, air-gapped environments, immutable audit logging, decentralized identity compatibility, and advanced policy enforcement through technologies such as Open Policy Agent, SPIFFE/SPIRE, WebAssembly isolation, and cryptographic verification systems. These capabilities are designed to help protect users from prompt injection attacks, unauthorized access, data leakage, hidden automation, and autonomous policy bypasses.
Covenant AI is also designed as a transparent and community-driven ecosystem. Licensed under the GNU Affero General Public License v3.0 or later (AGPL-3.0+), the project encourages open collaboration, public security review, and interoperable standards development. The platform includes support for modern AI infrastructure technologies including Rust, Python, TypeScript, Kubernetes, PostgreSQL, Qdrant, Ollama, vLLM, Hugging Face Transformers, and multiple open-source model ecosystems. Through its open governance approach and vendor-neutral architecture, Covenant AI aims to provide a long-term foundation for accountable, privacy-first, and human-authority-centered artificial intelligence systems.

- Covenant AI — An open-source AI infrastructure platform built around human authority, bounded autonomy, and fully verifiable, privacy-first execution.
