Chat with your community data · Lab

A chat app where a game's community team asks about its Discord, Reddit and Steam data and agent workflows answer; Barış Ülgen built the app, and my code was a few changes to the Discord collector.

What it is

FIG. 1From a server to an answer. The chat app only starts a job and polls for it; the workflows do the reading. My commits were on the Discord bot.
Source: cogniscope-ui README.md and src/services/chatService.ts; cogniscope-discord-bot; client-n8n workflow names (lab inventory B, row 5)

A chat app for a game's community team. You ask what players are saying on Discord, on Reddit or in Steam reviews, and agent workflows built in n8n answer from the collected data, with SQL for counts and retrieval for what people wrote. A switch moves between community analysis and market research. Long analyses run as jobs that the app checks every five seconds, which kept slow answers clear of a two-minute proxy limit.

What happened

The product moved in steps: a community dashboard, then agents that answer questions on demand (this app), then an AI community manager that could moderate, reply and open tickets. Barış Ülgen built the chat app, 94 of its 95 commits; the record does not show who designed which workflow. My own code here is three commits to the bot that collects a Discord server's messages in batches.

The closed beta had paying users; the first ones paid for access out of their own pockets.

We later retired the AI community manager, once general AI assistants could do what we had planned for it. We kept the daily collection of each game's community data, organised with embeddings.

What I learned

The beta user who left did not leave over the idea. The reason we wrote down was source attribution: people needed to see where an answer came from before they trusted it. A market review in April 2026 also found community insight crowded, with eight or more tools built for games and free insight tools from Discord and Reddit.