AI-Driven Project Management: Introducing MCP-Vikunja

In modern software engineering, AI can already write code. But writing code is only part of the development process. AI-Driven Project Management: GitHub Page MCP-Vikunja

Real engineering requires structure:

• Task tracking
• Iteration cycles
• Prioritization
• Status transitions
• Continuous refinement

At cn3m0, we asked a different question:

What happens when an AI agent can manage its own development workflow?

This led to the creation of mcp-vikunja — a Model Context Protocol (MCP) adapter that connects AI systems like OpenAI Codex to a self-hosted Kanban infrastructure powered by Vikunja.

The Problem: AI Without Structure

Most AI coding agents operate in a linear interaction model:

1. Receive prompt
2. Generate code
3. Return output

What is missing?

• Persistent project state
• Task memory
• Structured iteration
• Status awareness

Without a task system, the AI does not operate inside a development lifecycle. It reacts — it does not manage.

The Solution: MCP + Vikunja

Vikunja is an open-source, self-hostable project management platform providing:

• Lists
• Kanban boards
• Gantt views
• Table views
• Structured project organization

With mcp-vikunja, we expose Vikunja’s REST API through an MCP-compatible adapter layer.

This allows AI systems to:

• List projects
• Create new projects
• Create tasks
• Move tasks between Kanban buckets
• Read project state
• Generate structured summaries

All within a fully self-hosted environment.

Architecture Overview

The system is intentionally minimal:

AI Agent (Codex) → MCP Adapter → Vikunja REST API → PostgreSQL

The AI no longer operates purely in text space — it interacts with a structured task graph. This shifts the AI from reactive code generation to active process participation.

Why Self-Hosted Matters

Many AI workflows depend on cloud SaaS tools.

Our design principles for mcp-vikunja were:

• Fully self-hosted
• Docker-based deployment
• No external cloud dependencies
• Token-based authentication
• Deterministic behavior

This ensures:

• Data sovereignty
• Reproducibility
• Long-term stability
• Infrastructure independence

For serious AI engineering, this is not optional.

From PoC to Long-Term Experiment

The initial Proof of Concept has been completed successfully.

The system can:

• Spin up via Docker Compose
• Authenticate via API token
• Create and move tasks programmatically
• Reflect current board state

We are now moving the project into long-term testing.

The key research question:

Can AI agents not only write code — but manage structured development processes autonomously?

Practical Use Cases

Potential applications include:

• AI-managed feature roadmaps
• Iterative bug tracking
• Self-organizing coding sessions
• Autonomous task planning
• AI-generated weekly project summaries
• AI-assisted DevOps orchestration

This is the beginning of AI-native project infrastructure.

Repository

The project is open and available here: AI project management mcp-vikunja

Closing Thought

We are entering a phase where AI is no longer just a coding assistant. It is becoming part of the engineering loop. mcp-vikunja is a small but foundational step toward AI-native project orchestration. At cn3m0, we build systems where humans and AI collaborate inside structured infrastructure — not just in chat windows.