Self-Hosted AI Development Environment: Kandev on Netcup VPS
TL;DR: Running Kandev in 5 Minutes
Kandev is an AI-powered Kanban development environment that orchestrates multiple AI agents (Claude Code, Codex, GitHub Copilot, Gemini CLI, and more), manages tasks through visual workflows, and provides an integrated workspace for code editing, terminal access, and git changes review.
Quick Setup Commands:
# Prerequisites: Docker installed
npx kandev
Access the web UI at http://localhost:37429 and start orchestrating AI agents.
Recommended Netcup VPS:
- VPS 1000 G12 (4 vCore, 8 GB RAM, 256 GB NVMe) at EUR 10.37/month
- Use coupon 5799nc17718015261 for 50% off first month!
Introduction
Self-hosting AI development tools has become increasingly important for developers and teams who want full control over their workflow, data privacy, and customization options. Kandev represents a paradigm shift in how development teams can leverage multiple AI agents simultaneously while maintaining human oversight at every stage.
This comprehensive guide will walk you through understanding Kandev, its capabilities, and how to deploy it affordably on Netcup VPS infrastructure. Whether you are an individual developer or managing a team, this guide will help you set up a powerful AI-driven development environment without breaking the bank.
What is Kandev?
Kandev (AI Kanban & Development Environment) is an open-source platform designed to orchestrate multiple AI coding agents, manage tasks through Kanban-style workflows, and provide an integrated development workspace. Unlike traditional AI assistants that operate in isolation, Kandev enables parallel task execution, multi-agent collaboration, and structured review processes.
Core Features
Kandev offers a comprehensive set of features that transform how development teams work with AI:
Multi-Agent Support Kandev supports various AI agents including Claude Code, Codex, GitHub Copilot, Gemini CLI, Amp, Auggie, and OpenCode. This flexibility allows teams to choose the best agent for specific tasks or leverage multiple agents in a single workflow.
Parallel Task Execution Start and manage multiple tasks from different sources simultaneously. This capability significantly boosts productivity when working with AI agents, as you can delegate independent tasks to run concurrently rather than waiting for each to complete sequentially.
Integrated Workspace The platform provides a unified interface featuring a built-in terminal, code editor with Language Server Protocol (LSP) support, git changes panel, embedded VS Code, and chat functionality. All these tools work together in one IDE-like view, eliminating the need to switch between multiple applications.
Kanban Task Management Drag-and-drop boards with customizable columns and workflow automation enable teams to visualize and manage development tasks effectively. Task progress is clearly visible, and automation rules can trigger agent actions based on task state changes.
Agentic Workflows Multi-step pipelines define how agents work on tasks. For example, you can configure Claude Code Opus to design a plan, GitHub Copilot Sonnet to implement it, and Codex GPT 5.4 to review changes - all coordinated through a single task workflow.
Workspace Isolation Using git worktrees, Kandev prevents concurrent agents from conflicting with each other. Each agent operates in its own isolated workspace, ensuring that parallel executions do not interfere with one another.
Flexible Runtimes Run agents as local processes, in isolated Docker containers, or in remote executors like Sprites. This flexibility allows you to choose the right execution environment based on your infrastructure and security requirements.
Supported Agents
Kandev communicates with agents via ACP (Agent Client Protocol). Some agents support ACP natively, while others use ACP adapter packages:
| Agent | Protocol |
|---|---|
| Claude Code | ACP (@zed-industries/claude-agent-acp) |
| Codex | ACP (@zed-industries/codex-acp) |
| GitHub Copilot | ACP |
| Gemini CLI | ACP |
| Amp | ACP (amp-acp) |
| Auggie | ACP |
| OpenCode | ACP |
Default Workflows
Kandev ships with five pre-built workflow templates:
Kanban Workflow Classic kanban board with automated agent execution: Backlog -> In Progress -> Review -> Done. Ideal for straightforward tasks where you want to assign work, let the agent run, and review the result.
Plan & Build Two-phase workflow where the agent creates a plan for your review before implementing: Todo -> Plan -> Implementation -> Done. Best for features that benefit from upfront design.
Architecture Focused on design and architecture without implementation: Ideas -> Planning -> Review -> Approved. Useful for capturing architectural decisions before writing code.
Feature Dev Full development lifecycle with quality gates: Todo -> Spec -> Work -> Review -> QA -> PR -> CI Fixup -> Done. Includes TDD loops, automated review, and QA testing.
PR Review Automated code review workflow: Waiting -> Review -> Done. The agent reviews changed files and produces structured findings organized into BUG, IMPROVEMENT, NITPICK, and PERFORMANCE categories.
Benefits of Self-Hosting Kandev
Data Privacy and Sovereignty When you self-host Kandev, all your code, agent interactions, and development data remain on your infrastructure. This is critical for teams working with sensitive codebases or operating under strict compliance requirements.
No Vendor Lock-In Kandev is open-source (AGPL-3.0 license) with no telemetry. You are not dependent on any cloud provider's terms of service or pricing changes.
Full Control Configure your own authentication, network access, and security policies. Deploy behind your VPN or firewall for complete control over who can access your development environment.
Cost Efficiency Running Kandev on your own VPS eliminates per-user or per-seat licensing costs associated with commercial AI development tools.
Customization Extend and modify Kandev to fit your team's specific workflows. Add custom skills, MCP servers, and agent configurations tailored to your codebase and coding standards.
How to Use Kandev
Basic Concepts
Understanding Kandev's architecture helps you use it effectively:
Tasks and Workflows Tasks represent work items that AI agents will execute. Workflows define the stages a task moves through and what happens at each stage. When a task enters a workflow step, Kandev can automatically start an agent, wait for completion, or prompt for human input.
Agents and Executors Agents are the AI coding assistants (Claude Code, Codex, etc.) that perform actual work. Executors determine how and where agents run - locally, in Docker containers, or on remote infrastructure.
Workspaces and Worktrees Each task gets its own git worktree to prevent conflicts. Agents work in isolated directories, ensuring parallel tasks do not interfere with each other.
Workflow Execution
When you create a task and assign it to a workflow:
- Task Creation - You define the task description and select a workflow
- Automatic Transitions - Tasks move between workflow steps based on events (message sent, agent completed, etc.)
- Agent Execution - At designated steps, Kandev starts the configured agent automatically
- Human Review - You review the agent's output at review stages
- Completion - The task reaches the final state (Done or Approved)
Interacting with Agents
Kandev provides multiple ways to interact with agents:
Structured Chat - Send messages through the Kandev interface for structured communication CLI Passthrough - Drop into raw agent CLI mode for full terminal interaction with agent TUIs MCP Tools - Agents can call Kandev's MCP tools to save plans, designs, and artifacts
Quick Start Guide
This section provides detailed instructions to get Kandev running on your Netcup VPS in under 10 minutes.
Prerequisites
Before starting, ensure you have:
- A Netcup VPS (VPS 1000 G12 or higher recommended)
- SSH access to your server
- Basic knowledge of command-line operations
Step 1: Order Your Netcup VPS
- Visit https://www.netcup.com/en/server/vps
- Select VPS 1000 G12 (4 vCore, 8 GB RAM, 256 GB NVMe) at EUR 10.37/month
- Use coupon 5799nc17755868250 for an introductory discount
- Choose your preferred location (Vienna, Nuremberg, Amsterdam recommended for EU users)
- Select 12-month billing term for best pricing
- Complete registration and payment
Step 2: Initial Server Setup
After receiving your server credentials:
Connect to your VPS via SSH:
ssh root@your-server-ipUpdate system packages:
apt update && apt upgrade -y
Install prerequisites (Docker, Git, Node.js):
# Install Docker curl -fsSL https://get.docker.com | sh # Install Git apt install -y git # Install Node.js 18+ curl -fsSL https://deb.nodesource.com/setup_18.x | bash - apt install -y nodejs
Step 3: Install Kandev
Using the recommended NPX method:
# Run Kandev with NPX (downloads pre-built bundles)
npx kandev
For production deployment with persistent data:
# Clone the repository
git clone https://github.com/kdlbs/kandev.git /opt/kandev
cd /opt/kandev
# Start in production mode
make start
This starts both backend and frontend services. The worktrees and SQLite database are created in ~/.kandev by default.
Step 4: Access the Web Interface
After starting Kandev, access the web UI:
- Open your browser to http://your-server-ip:37429
- Or use SSH port forwarding:
ssh -L 37429:localhost:37429 root@your-server-ipand open localhost:37429
Configure your first agent (e.g., Claude Code):
- Install the agent CLI on your server
- Add agent credentials in Settings > Agents
- Test the connection
Step 5: Create Your First Task
- Click "New Task" in the Kandev interface
- Enter a task description (e.g., "Create a simple Python web scraper")
- Select a workflow (e.g., "Plan & Build")
- Click "Create"
- Watch as the agent analyzes the task and creates a plan
Running Kandev with Docker (Alternative)
For isolated container execution:
# Build the Docker image
docker build -t kandev:latest .
# Run Kandev in a container
docker run -d \
--name kandev \
-p 37429:37429 \
-v kandev-data:/root/.kandev \
kandev:latest
Configuration Options
Kandev can be configured through environment variables:
# Set custom port
KANDEV_PORT=37429
# Set data directory
KANDEV_DATA=/opt/kandev/data
# Enable HTTPS (reverse proxy required)
KANDEV_HOST=your-domain.com
Choosing the Right Netcup Server for Kandev
Selecting the appropriate VPS for Kandev depends on your usage patterns, team size, and budget. Here are Netcup's current VPS Generation 12 offerings:
VPS Comparison for Kandev
| Plan | CPU | RAM | Storage | Monthly Price | Best For |
|---|---|---|---|---|---|
| VPS 500 G12 | 2 vCore | 4 GB | 128 GB NVMe | EUR 5.91 | Personal use, testing |
| VPS 1000 G12 | 4 vCore | 8 GB | 256 GB NVMe | EUR 10.37 | Small teams, 2-3 users |
| VPS 2000 G12 | 8 vCore | 16 GB | 512 GB NVMe | EUR 19.25 | Medium teams, parallel agents |
| VPS 4000 G12 | 12 vCore | 32 GB | 1024 GB NVMe | EUR 32.41 | Large teams, complex workflows |
| VPS 8000 G12 | 16 vCore | 64 GB | 2048 GB NVMe | EUR 47.95 | Enterprise, heavy workloads |
Recommended Options
Personal Use / Testing (VPS 500 G12)
- Price: EUR 5.91/month
- Coupon: 5799nc17755868070
- Use for: Learning Kandev, testing workflows, personal projects
Small Teams (VPS 1000 G12)
- Price: EUR 10.37/month
- Coupon: 5799nc17755752870
- Use for: Running 2-3 concurrent agent tasks, day-to-day development
Growing Teams (VPS 2000 G12)
- Price: EUR 19.25/month
- Coupon: 5800nc17718015232
- Use for: Multiple developers, parallel workflow execution, database storage
Professional Development (VPS 4000 G12)
- Price: EUR 32.41/month
- Coupon: 5801nc17718015213
- Use for: Complex multi-step workflows, CI/CD integration, large codebases
Resource Considerations
CPU Kandev's backend and frontend have modest CPU requirements. However, running AI agents in Docker containers can be CPU-intensive. More vCPUs allow parallel agent execution without performance degradation.
RAM 8 GB minimum recommended for running one agent with context. For parallel execution or multiple concurrent tasks, consider 16 GB or more.
Storage Kandev stores worktrees and SQLite database locally. Factor in your codebase sizes when selecting storage. The default 256 GB on VPS 1000 G12 accommodates most projects.
Cost Optimization Tips
- Use Coupons - Always apply available coupons for first-time orders
- Hourly Billing - For testing, use hourly billing (EUR 0.017/hour for VPS 1000) to minimize costs
- 12-Month Term - Commit to 12 months for lower monthly rates
- Scale Up Later - Start with a smaller plan and upgrade as needed
Coupons Summary
| Server Plan | Coupon Code | Offer |
|---|---|---|
| VPS 500 G12 | 5799nc17755736850 | 50% off first month |
| VPS 1000 G12 | 5799nc17755735170 | 50% off first month |
| VPS 2000 G12 | 5800nc17718015231 | 50% off first month |
| VPS 4000 G12 | 5801nc17718015212 | 50% off first month |
| VPS 8000 G12 | 5802nc17718015174 | 50% off first month |
Use coupon 36nc17718015546 for EUR 5 off any order!
Conclusion
Kandev represents a powerful, self-hostable solution for teams looking to orchestrate multiple AI coding agents within a structured development workflow. By running Kandev on Netcup VPS, you gain complete control over your development environment while benefiting from Europe's best VPS pricing.
Key Benefits Recap:
- Full Data Sovereignty - Your code and AI interactions stay on your infrastructure
- Multi-Agent Orchestration - Use Claude Code, Codex, Copilot, and more in parallel
- Structured Workflows - Kanban, Plan & Build, Feature Dev, and custom pipelines
- No Vendor Lock-In - Open source, AGPL-3.0 licensed, no telemetry
- Cost Effective - Starting at EUR 5.91/month with coupons
Getting Started:
- Order a Netcup VPS (VPS 1000 G12 recommended for teams)
- Apply coupon 5799nc17755734540 at checkout
- Deploy Kandev using
npx kandev - Configure your first AI agent
- Start orchestrating development tasks
With prices starting at just EUR 5.91 per month and coupon discounts available, Kandev on Netcup VPS offers enterprise-grade AI development capabilities at a fraction of commercial alternatives. The combination of flexible runtimes, multi-agent support, and visual workflow management makes Kandev an essential tool for modern development teams embracing AI-assisted coding.