yangheng95/opencorvus

DIY your own agent team harness for long-horizon agent work

What it solves

OpenCorvus is designed for long-horizon, complex agent tasks that typically fail when handled by a single prompt or a simple loop. It provides a structured harness to ensure that research, implementation, and review processes are connected and that progress is retained across interruptions, allowing for the completion of high-complexity goals with verifiable results.

How it works

The system uses "expert squads"—coordinated groups of AI agents with specific roles and responsibilities. These squads can be manually selected or dynamically composed based on the task's requirements. The runtime manages the coordination of these agents using scheduling contracts that combine Directed Acyclic Graphs (DAGs), local refinement loops, and conditional branches to move a task forward.

Who it’s for

It is intended for developers and researchers who need to automate complex, multi-stage workflows (such as writing a research paper or developing a game) that require independent review and a persistent record of tool calls and artifacts.

Highlights

  • Expert Squads: Supports both pre-defined and dynamically composed agent teams.
  • Long-Horizon Support: Maintains task state and context to allow recovery after interruptions.
  • Verifiable Results: Keeps a detailed work record of tool calls, artifacts, and review evidence.
  • Broad Integration: Integrates with 87 model providers and supports the Model Context Protocol (MCP).
  • Desktop App: Provides a user-friendly interface for submitting goals and acceptance criteria.

Related

  • Project
    RizRiyz/luvus

    Luvus is a cross‑platform terminal UI that lets you launch, monitor, and orchestrate AI coding agents (Claude, Codex, Grok, etc.) alongside shells, editors, and Git tools. It provides persistent workspaces, pane management, agent‑aware status panels, Git/GitHub integration, remote‑machine support, a read‑only web view, and an extensible module system. Install via a script, Homebrew, or PowerShell, then use `luvus` to start a session, `luvus doctor` to verify setup, and `luvus update` for upgrades.

  • Project
    different-ai/openwork

    OpenWork is an open‑source, cross‑platform desktop app that lets AI agents work on your local files. It supports any LLM (hosted APIs, custom keys, or local Ollama models) and provides a Model‑Control‑Plane (MCP) with tools for discovering and executing shared skills. Teams can manage access, publish plugins, and self‑host the optional OpenWork Den control plane. The core app is MIT‑licensed; the Den component is source‑available under an EE license (free for small teams, paid for larger orgs).

  • Project
    GCWing/OpenBitFun

    OpenBitFun is a cross-device, highly customizable agent workspace that allows users to manage, extend, and run AI agents across desktop, mobile, and remote environments.

  • Project
    synthetic-sciences/openscience

    OpenScience is an open‑source AI research workbench that lets users describe scientific tasks in plain language. The built‑in agent plans, searches literature, writes and runs code (Python/R/shell), executes experiments on local or remote compute, and produces a reproducible report, with every step logged. Available as a desktop app, web workspace, or CLI, it supports custom LLMs, a large library of domain‑specific skills, and extensible plugins.