OpenCorvus
An open-source harness for long, complex agent tasks that coordinates squads of expert agents, tracking task state so work can be checked, resumed after interruption, and carried forward across research, implementation and review.
OpenCorvus is an open-source harness for long-horizon agent work, bringing models, tools, context and expert-squad coordination into one runtime rather than one large single-agent task.
Work moves through connected roles, such as research, implementation and independent review, with clear responsibilities. Task state and context are retained so work can continue or recover after an interruption, and tool calls, artifacts and review evidence stay in the work record so the result can be checked afterward, rather than only trusted.
It is used through a desktop app, choosing a reachable model provider, a project and a squad, then submitting a goal and acceptance criteria, or built from source with Bun. It evolved from the OpenCode codebase and still shares some of its model-provider and GitHub Copilot integration code.