Brain discovery: how AI assistants find the right knowledge brain

Brain discovery is how an AI assistant finds expert knowledge brains it can recommend, beyond the ones you already subscribe to. In KBrain it is powered by a dedicated MCP tool, discover_brains, that searches the marketplace during a conversation and returns the brains that best match your question - each with a link you can open to subscribe. This article explains what it does, when it runs, and why it matters for both people asking questions and people publishing knowledge.

Two different jobs: answering vs. recommending

KBrain gives an assistant two distinct capabilities, and keeping them separate is the whole idea. search_knowledge answers from a brain you are connected to - it reads the curated content and grounds the reply in it. discover_brains does not read content at all; it looks at the public catalog and recommends a brain you could subscribe to when your connected brains fall short.

How discover_brains works

When you connect KBrain to Claude or ChatGPT over MCP, the assistant sees discover_brains alongside its other tools. If it decides your question is not covered by your connected brains, it calls the tool with your topic and, optionally, a category. KBrain searches published marketplace brains, ranks them by relevance and popularity, excludes anything you already subscribe to, and returns a short list.

• Matches on brain name, description, tags, and category so results fit the question
• Excludes brains you already have, so it only surfaces something new
• Returns a direct subscribe link for each result, ready to open after login
• Reads only public catalog metadata - never the private contents of a brain
• Runs the same way in Claude, ChatGPT, and any MCP-compatible assistant

When discovery runs - and when it stays quiet

Discovery is gap-triggered by design. The assistant is instructed to reach for discover_brains only when its connected brains do not cover your topic - not on every message. That keeps recommendations useful and stops the experience from turning into a constant upsell. If your subscribed brains already answer the question, discovery never fires.

Why it matters for people asking questions

The hardest part of getting a reliable answer is knowing that a better source exists. Discovery removes that step. Instead of an assistant confidently guessing on a topic it was never trained on, it can tell you exactly which expert brain would answer it - and get you there in one click. That means fewer hallucinations and more sourced, trustworthy answers.

Why it matters for brain creators

For anyone publishing knowledge, discovery is distribution. A brain no longer has to wait to be browsed - it can be recommended the instant a user asks a question it answers. Relevance to the question, not marketing budget, is what surfaces your brain, and the subscribe link turns that relevance into real subscribers inside the assistants people already use.