ABOUT
Built by one person who needed it to exist.
I'm Greg. I build SaaS products. And for years I watched the same thing happen: people would sign up, land on the dashboard, stare at it for 30 seconds, and leave. They never came back.
I tried the existing tools. Pendo wanted $7,000/year for a product making $200/month. WalkMe needed an "implementation specialist." Appcues made me manually author every tooltip flow, which broke every time I shipped a UI change. I was spending more time maintaining onboarding than building features.
So I built the thing I wished existed: an AI agent that reads your app's DOM — buttons, tabs, forms, disabled states — and guides users automatically. No manual flow authoring. No CSS selectors that break on deploy. No $50K enterprise contracts.
One script tag. The AI does the rest.
Why solo?
ActivationKit is intentionally a one-person company. No investors, no board, no pressure to optimize for metrics that don't matter. Every decision is made for one reason: does this help users get unstuck faster?
The AI is the team. It handles what would normally require a customer success department: answering questions, surfacing features, generating tours, detecting when users are confused. That's the whole point — if the product needs a team to run, it's not good enough.
The principles
Observe the DOM, never require host cooperation
Every feature works by reading what's already on the page. Disabled buttons, hidden elements, active tabs — the AI sees it all. Your app doesn't need to configure anything or pass data. If the AI can't figure it out from the DOM, that's our bug, not your setup task.
Free means free
25 AI conversations per month, forever. Not a 14-day trial that pressures you into paying before you've experienced the value. Use it until you outgrow free. Upgrade because you want to, not because a timer ran out.
Ship fast, fix faster
ActivationKit runs on its own product. Every bug I find as a user, I fix as the builder. Pulse alerts notify me in Slack the moment something goes wrong — before any customer has to report it.
The stack
TypeScript monorepo. Hono API on Fly.io. Next.js dashboard on Vercel. Neon Postgres. Shadow DOM SDK at 30KB gzipped. Three-tier decision pipeline: rules (<10ms), cache (<50ms), LLM (when it actually needs to think). Total infrastructure cost: ~$2/month.
The architecture is documented publicly. The code is proprietary because the AI logic is the product, but the approach is transparent. If you want to know how something works, ask — I'll tell you.
Questions? Feedback? Just want to talk shop?