Managing projects inside ClickUp involves a lot of small, repetitive manual actions: creating tasks one by one, digging through nested folder/list structures, checking on space and project lists. None of it is hard individually, but it's exactly the kind of fragmented, click-heavy work that eats time without ever feeling like real work getting done.

An AI co-agent, built on LangGraph and Gemini, that sits between the user and ClickUp's API. Instead of navigating ClickUp's UI, the user describes what they want in plain language, and the agent carries it out directly against ClickUp: task creation, fetching project/space lists, creating lists and folders, and retrieving the full nested list structure on request.

LangGraph structures the agent's reasoning and multi-step actions; Gemini handles the natural-language understanding layer. The agent also used AG-UI — a protocol, genuinely new at the time, that lets an agent manipulate the UI directly rather than just returning text — an early-adoption, technical-credibility signal in its own right.

The interesting design decision wasn't "add a chatbot to ClickUp" — it's giving the agent enough structural awareness of ClickUp's own hierarchy that it can act on the platform's actual organization, not just create flat, disconnected tasks.