Coven: Local-First Runtime for AI Coding Agent Sessions

Summary

Coven is a powerful local-first runtime designed for managing project-scoped AI coding agent sessions. It provides durable state, robust authority boundaries, and seamless interoperability with multiple coding harnesses like Codex, Claude Code, and GitHub Copilot CLI. This tool enables developers to safely launch, observe, and coordinate AI agent work directly within their local project environments.

Use at your own risk

OSRepos shares public repositories for knowledge and discovery only. Any installation, execution, configuration, or use of code from these repositories is the user's own responsibility. Always review the repository, source code, dependencies, licenses, and security implications before running or installing anything. OSRepos is not responsible for issues, damages, or losses resulting from third-party repositories.

Introduction

Coven is a cutting-edge local-first runtime developed by OpenCoven, specifically engineered for project-scoped AI coding agent sessions. It acts as a neutral substrate, allowing developers to integrate and manage various coding agent CLIs, including Codex, Claude Code, and GitHub Copilot CLI, within explicit local project boundaries. Coven ensures durable state, establishes clear authority boundaries, and facilitates multi-harness interoperability, making it a pivotal tool for modern AI-assisted development workflows.

Why Use Coven and Its Benefits

Coven addresses the need for a controlled and visible environment for AI coding agents. Its core philosophy is "One project. Any harness. Visible work." This means you gain unparalleled flexibility and control over your AI-driven development:

Harness Flexibility: Choose your preferred coding agent, whether it's Codex, Claude Code, GitHub Copilot CLI, or future adapters, and run them all through a unified system.
Session Ownership: Coven takes ownership of the agent session, providing project-scoped boundaries, secure PTY execution, comprehensive event logging, and SQLite persistence. This ensures that agent activities are contained, auditable, and their state is durable.
Visible Work: All agent interactions and outputs are made visible, allowing developers to observe, attach to, and coordinate agent work effectively.
Security and Control: The Rust daemon acts as the authority boundary, ensuring that security decisions flow inward, enhancing the safety of your development environment.
Local-First Development: By operating locally, Coven offers speed, privacy, and reliability, reducing dependency on external services for core agent operations.

Installation

Getting started with Coven is straightforward using npm. Ensure you have Node.js installed, then run:

npm install -g @opencoven/cli
coven doctor

Coven's universal wrapper (@opencoven/cli) automatically selects the correct platform-specific package for your system (macOS Apple Silicon, macOS Intel x64, glibc-based Linux x64, or Windows x64). For more detailed installation routes, including cargo and source builds, refer to the official documentation.

Examples

Here's a quick guide to launching and managing your first AI coding agent session with Coven:

cd /path/to/your/project

# 1. Complete provider-owned login
coven setup codex

# 2. Check local readiness
coven doctor

# 3. Start the daemon
coven daemon start

# 4. Launch a session
coven run codex "fix the failing tests"

# 5. Browse and manage sessions
coven sessions

# 6. Stop the daemon when done
coven daemon stop

Running coven without arguments opens the interactive Coven UI, providing a visual interface for managing your sessions. The full command reference is available in the CLI documentation.