Mng-dev-ai/agentrove

What it solves

Agentrove provides a unified, self-hosted workspace for managing and orchestrating multiple AI coding agents (such as Claude Code, Cursor, and Grok) from a single interface. It eliminates the need to switch between different agent tools by providing a centralized hub for chat, code editing, and terminal access, while ensuring each workspace is isolated in its own sandbox.

How it works

The system uses ACP adapters to run various AI agents. Each workspace is hosted in a Docker or host sandbox, integrating a code editor, terminal, file tree, and git tools.

Crucially, it includes a bundled MCP server that allows agents to act as orchestrators. A "lead" agent can spawn worker sub-threads—each running on a different model or persona—to handle specific tasks in parallel. These workers can operate in isolated git worktrees to avoid conflicts, and the lead agent reviews their output before finalizing the result.

Who it’s for

Developers who use multiple AI coding assistants and want a centralized, self-hosted environment to manage complex coding tasks through multi-agent orchestration.

Highlights

• Multi-Agent Orchestration: A lead agent can decompose tasks and delegate them to specialized worker agents in parallel sub-threads.
• Integrated Workspace: Combines chat, a Monaco-based code editor, terminal, and git tools in one view.
• Sandbox Isolation: Each workspace runs in its own Docker or host sandbox for security and isolation.
• Crosspatform Availability: Available as a Docker web app, a macOS desktop app (via Tauri), and a native iOS app.
• Git Integration: Supports GitHub-assisted repository browsing, PR creation, and branch management.