Built by an idea
and Claude Code.
Feeling HQ was not built the traditional way. There was no engineering team, no design sprint, no sprint backlog. There was a conversation — and then there was a product. Almost every line of code, every component, every deployment pipeline was written and maintained by Claude.
A thought, typed into a chat window.
It started the way most things do: with a problem. Running a meeting with fifty people and needing a quick, honest read on the room. A feelings wheel printed on paper just wouldn't scale. So the question was simply asked — "could we build this as a web app?" — and Claude said yes.
Architecture before a single line of code.
Rather than diving straight in, the conversation shifted to design. Which stack? How should the wheel render? What's the data model for nested emotions? Claude mapped out the component tree, proposed Next.js with a Convex backend for real-time responses, and sketched the API surface. It read like a senior engineer's RFC — because, in a way, it was.
Claude Code wrote the product, file by file.
With a plan in place, Claude Code took over the keyboard — often in cowork mode, where human and AI work side by side in the same session, the human steering direction while Claude handles every keystroke. The interactive SVG wheel, the live word-cloud for group events, the AI feeling-matcher, the dark mode system, the design tokens, the responsive layout — each piece landed clean and connected. When a component broke, Claude diagnosed and fixed it. When a design looked off, Claude iterated. The human role was vision; the execution was almost entirely machine.
Agents that never clock out.
After launch, a fleet of specialised Claude agents took over ongoing operations. One watches Search Console and adjusts SEO. Another writes and publishes content. A third monitors the repo, reviews diffs, and manages deployments through CI/CD. All of this runs through Claude Desktop, which has access to the broader desktop environment — browsers, terminals, file systems — enabling agents to act with the same reach a human operator would have.
After launch, a set of specialised Claude agents — operating through Claude Desktop with full desktop access — keep the product alive, optimised, and growing.
SEO Agent
Monitors Google Search Console, analyses ranking opportunities, rewrites meta descriptions, and submits sitemaps — all autonomously.
Content Agent
Gets commissioned via a Slack message — even when the human isn't around. It researches topics, writes new blog posts, and publishes content end-to-end without any further input.
Analytics Agent
Reads usage patterns, surfaces what's working and what isn't, then proposes data-driven improvements to the UX.
Repo & CI/CD Agent
Reviews pull requests, manages branches, triggers deployments, and monitors build health — the full DevOps loop, unattended.
Built entirely with Claude Code
From the first component to the last deployment hook, Claude Code authored the code. The human provided intent and direction — Claude provided execution. This page included.
"The most interesting thing about Feeling HQ isn't what it does — it's how it exists. A product built, deployed, and maintained almost entirely by AI, serving real people with real emotional needs."
— The human behind the idea
Try it yourself.
Explore the feeling wheels, run an event with your team, or just see how AI-built products feel when used by humans.