TaskFlow

TaskFlow is a production-ready task management platform born from a different question than usual: what happens when the main users are no longer only people, but also AI agents? Traditional project management assumes someone clicks, drags and comments.

The product covers everything you expect from such a tool — a Kanban board with drag-and-drop and optimistic UI, full project and task CRUD, nested comments — but with a twist: an AI-first API layer at /api/v1/context designed so agents like OpenClaw can read context and act autonomously.

Actions are therefore available through two channels: via the interface, optimistically and immediately, and via the API, programmatically. Agents use granular-permission API keys, with SHA-256 hashing and defined scopes, so automation is not full access but a controlled delegation.

I built the app on Next.js 14, React, TypeScript and Supabase, with Zod for input validation and Framer Motion for transitions. The interesting challenge was making a reactive interface for humans coexist with a stable API contract for machines.

TaskFlow moves project management from the human panel to the agent: the same flows can be governed by a person or orchestrated by automation, without changing the data model. It is a concrete experiment in what software becomes when users are hybrid.