AI Agent

TaskGenAgent for Agile Task Management

TaskGenAgent is an AI agent for task generation within the HIVEMIND project , designed to break down software specifications into prioritized, structured, and actionable development tasks. As part of a human-centred LLM-based multi-agent system for software engineering, TaskGenAgent supports Agile task management by acting as a bridge between software requirements, architectural decisions, project artefacts, and project-management workflows.

Its goal is to transform high-level software knowledge into clearer task definitions enriched with planning information, including priorities, effort estimates, dependencies, scheduling details, required roles or expertise, resource allocation, and traceability links to the original requirements or architectural inputs.

Research team

TaskGenAgent is being developed at Universitat Politècnica de Catalunya (UPC) within the HIVEMIND. The UPC team involved in this line of work includes Carles Farré, Lidia López, Xavier Franch, and Razieh Akbari. The work focuses on defining and developing human-centred AI support for software task generation, requirements-to-task generation, and structured Agile planning.

In this context, the agent contributes to the broader HIVEMIND vision of collaborative multi-agent systems, where specialised AI agents support different stages of the software development lifecycle and help software teams move from requirements and design knowledge toward traceable, actionable, and reviewable development tasks.

Talks and updates

Progress on this research has been presented at several venues, including GAISE 2026 in Tampere, Finland, and the HIVEMIND M18 General Assembly in Bucharest, Romania. Read more about these presentations and ongoing updates below:

• Presenting TaskGenAgent at GAISE 2026
• TaskGenAgent at the HIVEMIND M18 General Assembly
• Generative UI for TaskGenAgent
• More about my research background

×