kdlbs/kandev

What it solves

Kandev provides a centralized control plane for managing and orchestrating multiple AI coding agents. It solves the problem of scaling agentic workflows by allowing users to run tasks in parallel, offload execution to remote servers, and review agent-generated changes in a single integrated workspace rather than relying on fragmented terminal-based interfaces.

How it works

Kandev acts as an orchestrator that connects to various AI agent CLIs (via the Agent Client Protocol or raw TUI passthrough). It manages the lifecycle of these agents across different runtimes—including local processes, Docker containers, SSH remotes, and cloud environments. To prevent conflicts, it uses Git worktrees to isolate each agent's workspace. It also provides a web-based UI with a kanban board for task management and an IDE-like workspace featuring a code editor, terminal, and git changes panel.

Who it’s for

It is designed for power users and development teams who want deep control over their AI agent workflows, prefer self-hosted infrastructure, and need to orchestrate complex, multi-step pipelines across multiple repositories.

Highlights

• Multi-agent support: Integrates with a wide array of providers including Claude Code, GitHub Copilot, Gemini, and Devin.
• Parallel execution: Run multiple agent tasks simultaneously across different sources.
• Agentic workflows: Create multi-step pipelines where different agents are assigned to different steps (e.g., one for planning, one for implementation, one for review).
• Remote runtimes: Offload agent execution to remote servers via SSH or Docker to avoid saturating local machine resources.
• Integrated workspace: A unified view combining a file editor, terminal, browser preview, and git changes panel.
• Multi-repository support: Span a single task across multiple repositories with dedicated worktrees and branches per repo.