TaskCycle began as an internal tool for my own commercial software work. Before building it, I worked as a developer, solution architect, team leader, and project manager. I had structured teams to create, test, deploy, and support software. I wanted to bring those same principles to working with AI agents.
The video above is the original 2:54 founder introduction. For the organizational challenges discussed in workshops with software-house owners and CTOs, watch the newer explanation and read the article.
Why an agent team?
Rather than asking one general-purpose assistant to do everything, I organized the work around specialized roles. The agent team has distinct responsibilities, context, system prompts, models, and tool access. Those boundaries matter when work continues for hours: agents need the right information and tools without requiring a person to supervise every step.
Control is part of the workflow
A development task is not done just because an agent says it is. Command policy and the Security Gate add controls around execution and check completed changes with validation tools. Their findings and evidence remain available to the reviewer. These checks reduce risk; they do not guarantee that software is secure or replace professional code review.
Human review is the destination
My aim is to let the team do the intermediate work while keeping a qualified human responsible for reviewing the resulting code and evidence, giving feedback, and deciding whether to accept the change. That is what “human at the end of the cycle” means in practice.
I first built this for my own use, then saw that other teams might benefit from the same structure. If you are working out how to use agents across your organization, tell me about your workflow and join the private preview.