You've probably noticed every app looks like this now.
Let me explain why, and why chat isn't the real breakthrough.
Chat was the natural first interface for LLMs.
Easy for AI to understand, easy for users to interact with.
So, naturally, as ChatGPT / Anthropic / etc crossed billions in revenue and hundreds of millions of users, most other apps followed suit.
Even our experience is chat-first.
200 customers with hundreds of millions of users, all asking techical product questions through AI chat.
But what they really wanted was the outcome!
Think about it. No one logs into Salesforce to use Salesforce...they log in to grow their business.
We saw the same usage pattern across Kapa.
Chats like "How do I integrate Google Drive?" are actually asking "Can you integrate my Google Drive instance for me?"
This is where the puck is heading, and why companies like PostHog, Linear, and Notion are following in Claude / ChatGPT's footsteps.
Believe it or not, building this infrastructure is brutal.
To achieve this, you must focus on what happens after the chat request. Harness engineering, real-time RAG, orchestration, user-scoped auth, PII handling...
But that's the user unlock. An in-product assistant. They don't have to leave, and the assistant has access to your docs, their tools, and knows how to navigate both.
That's what the
@kapa_ai Agent Framework finally does.
The Kapa Agent Framework. Give it your tools and actions, connect one auth endpoint, and your users get a full in-product agent.
Port, Matillion, and Medusa already use kapa as part of their product agents, and we're onboarding more Kapa customers every day.