So this one is a bit different from the other experiments.
Its a tool I built for the brainylab team to think way more deeply about every prospect before we talk to them.
The goal of the tool is to give context and open ideas for diagnose opportunities.
When you walk into a first conversation with a potential client, the quality of that conversation is everything. If you show up with the angle, the story, the gaps, the context their competitors are filling — that changes the room. Generic talking points don’t.
So I built the brainylab Prospect App with Claude Code. Hosted locally. Not a product, just an internal tool our team feeds every time we take a new call.
![brainylab Prospect App — the input form that kicks off a multi-agent research pipeline]
The form is the front door. The more information the better.
Behind the form is where it gets interesting
The app is a multi-agent system.
Each agent handles one slice of the research properly, then passes its work forward.
- Agent 1 analyses the company website and their Google and Meta ads — messaging, positioning, what they’re actively pushing into the market
- Agent 2 reads the annual report and pulls financial and strategic signals
- Agent 3 researches key competitors — local players and the best global ones — and maps where the company sits
- Each of the three agents compiles its findings into its own research file
- Agent 4 agent reads all three files and writes the final document.
And this is the demo result for Nike. Its quite different from our actual research docs (because of the data, info, local specifics, benchmarks …), but you’ll get the idea.
As said this is just internal document to work on and the refine with the team before its sent out to the lead.
What I’m still iterating on is the competitor agent. Picking the right “best global competitor” is a judgment call, and sometimes it picks something adjacent but not really comparable. The next version will let us seed a couple of competitors manually to anchor that agent before it goes hunting.
Again I used my now quite refined AI agents flow proces. I wrote about it recently, and I’ve been building also live sites with it.
See brainylabs new product page, game library and framework landing. All of them were built with Claude Code AI agents process i wrote about here.
Access and use agents – see listing on Github.
