autonomous-ai/openharness

The ultimate harness for coding agents and beyond. All your agents. All your machines. One command center. Start with code, then follow your curiosity and build across disciplines: CAD, circuits, robots, games and music.

What it solves

OpenHarness provides a unified command center for managing multiple AI coding agents (such as Claude Code, Codex, and Cursor) across different machines (laptops, home servers, GPU boxes) in a single window. It eliminates the need for manual SSH keys, port forwarding, or complex network configurations while ensuring that agent activity remains end-to-end encrypted.

How it works

The system consists of a native app, a daemon that runs agents in tmux panes, and a relay for connectivity. It uses WebRTC for direct peer-to-peer connections between machines, falling back to Cloudflare TURN or a proprietary relay when firewalls block direct paths.

A "harness" is a specific running session of an agent with its own context. Users can create "Domain-Specific Harnesses" (DSH) which combine an agent with a specific toolchain, instructions, and a live viewer to turn a general coding agent into a specialist for fields like CAD, robotics, or music production.

Who it’s for

Developers and polymaths who use multiple AI agents and remote compute resources and want a high-performance, keyboard-first interface to orchestrate them.

Highlights

  • Multi-Agent/Multi-Machine Orchestration: Run and monitor various agents across different hardware in one interface.
  • End-to-End Encryption: All code and keystrokes are sealed using ChaCha20-Poly1305 and X25519 session keys.
  • Domain-Specific Harnesses: Extensible framework to integrate agents with specialized tools (e.g., Blender, MuJoCo, RDKit) and live viewers.
  • Native Performance: Built as a native app (not Electron) for low latency and only minimal CPU/RAM usage.
  • Open Hardware Integration: Support for a dedicated physical device (round AMOLED display) to monitor agents and provide voice input.
  • Worktree Isolation: Each harness starts an agent on its own branch to keep the main working copy clean.

Related

  • Project
    BennyKok/omg.dev

    omg.dev is an open‑source platform that lets you run coding‑assistant agents (Claude Code, Codex, Copilot, etc.) on your own computer or in the cloud, start tasks from a mobile app, and receive notifications when they finish.

  • Project
    giuliastro/harness-remote

    Harness Remote is an open‑source, local‑first control plane that lets you observe, pause, resume, and hand‑off native AI coding‑agent sessions (Codex, Claude, OpenCode, etc.) across devices. It runs a small Node.js gateway on a machine with a coding‑agent CLI, then connects via desktop, web, or Android clients. All code, credentials, and session data stay on your own machines; the tool only adds a UI and coordination layer.

  • Project
    Ark0N/Codeman

    A self-hosted mission control dashboard that manages multiple AI coding agents in persistent tmux sessions, providing a mobile-optimized web interface for remote monitoring and control.

  • Project
    Station-Sciences/bot-crossing

    A visual 3D dashboard that organizes AI coding agent threads from various tools into a spatial colony, allowing users to monitor and launch agent sessions across multiple repositories.

  • Project
    Aisland-SJL/dsh-worktable

    An agent-project workbench for DeepSeek Harness that provides a dockable multi-window workspace and a real-time control room for monitoring AI projects.