TaskFlow AI

Independent full-stack project · Apr. 2026

Visual task orchestration with human-in-the-loop automation

TaskFlow AI turns personal work into a visual system of tasks, dependencies, deadlines, and executable sub-workflows. It combines a node-based workspace with an orchestration boundary for OpenClaw and Lark.

TaskFlow AI dashboard showing tasks connected across workflow stages — The main canvas connects tasks to a status pipeline while keeping task context visible in a side panel.

InterfaceNested node flows

BackendFastAPI + SQLite

Agent layerOpenClaw + Lark

Why I built it

Conventional task lists flatten complex work into isolated rows. TaskFlow uses two levels of graphs: a personal canvas for major tasks, and an internal workflow for the serial, parallel, or merged steps inside each task. Status, deadlines, details, and completion rules remain attached to the graph.

Human control, agent assistance

OpenClaw acts as the orchestration layer: it can extract a candidate task from email or chat, propose structured details through Lark, and advance work when a user confirms the action or when evidence satisfies an explicit rule. TaskFlow remains the deterministic system of record instead of pretending that the product itself reads private accounts.

Internal TaskFlow workflow with serial and parallel sub-nodes — Each task can expand into an editable internal workflow.

TaskFlow analytics page with completion and risk summaries — Analytics surfaces workload, completion, and deadline risk.