WYKRA: Web You Know, Real-time Analysis

#devchallenge #n8nbrightdatachallenge #ai #webdev

n8n and Bright Challenge: Unstoppable Workflow

This is a submission for the AI Agents Challenge powered by n8n and Bright Data

What We Built

So you're tasked with finding influencers. Not just random ones, but ones that fit your niche, your vibe and maybe even your region. You type vague queries into Google. You scroll through Instagram. You open 19 Chrome tabs. You give up.

We built WYKRA so you don’t have to.

WYKRA stands for Web You Know: Real-time Analysis - a Telegram-based AI agent that connects real-time web scraping with natural language prompts to help you find or analyze Instagram influencers in seconds. Right now it only works with Instagram, but more social platforms are coming - stay tuned!

It solves two common problems:

“Find me influencers like...” – Give it a prompt, it gives you a list of matching Instagram profiles.

“Tell me more about this influencer” – Send it a link, it scrapes their profile and gives you a summary.

It’s powered by:

n8n
for the orchestration,

Bright Data
for real-time web scraping,

Neon
for storage and memory,

Telegram for interaction (no fancy frontend, just chat like a human).

Demo

Try the bot: https://t.me/wykra_bot

Demo videos:

Search for influencers using natural language:

Peek behind the profile:

n8n Workflow

Here’s our full workflow JSON in a GitHub Gist:
👉 https://gist.github.com/iscander01/f407ba565d22dc7bad46af37458657ee

Technical Implementation

Under the hood WYKRA is a multi-step orchestration between several components.

Model: We use Claude Sonnet for two key tasks:

Prompt classification - understanding whether the user wants to search or analyze.
Post-scraping summarization - turning raw Instagram profile data into concise, readable summaries.

User Input: Telegram messages are routed into n8n via the Bot API.

Prompt Flow: Claude Sonnet determines the user’s intent. If it’s a discovery request, it crafts queries; if it’s a profile link, it prepares for analysis.

Bright Data Web Scraper Node:

If the user provides an Instagram account, we use the Bright Data Verified Scraper (scrape-by-url) to extract public profile data.
If the user describes what kind of influencer they want, we interpret the prompt and search for relevant Instagram profile links, then analyze them one by one.

Neon DB stores both recent prompts and influencer profiles to improve performance and allow future caching.

n8n coordinates all of this in a visual workflow with conditional paths, retries and error handling.

We built two parallel flows inside one unified workflow, so one agent handles both discovery and analysis.

Bright Data Verified Node

We used Bright Data’s web scraper by URL node to extract data from public Instagram profiles. This gave us structured access to:

Number of followers/following
Bio and hashtags
Post count
Engagement signals

This scraping power is what makes WYKRA “real-time” - without Bright Data, we’d be stuck with stale or partial data.

Journey

We started with one goal: build a small but powerful AI agent that actually does something useful.

We knew we didn’t want yet another dashboard. We wanted something frictionless - so we picked Telegram. Just type what you want and get results.

Biggest technical hurdles:

Figuring out how to route two very different flows (search vs analysis) in the same workflow
Keeping the scraping node from failing on rate limits - Bright Data’s built-in retry and proxy rotation really helped here
Making results readable in Telegram (formatting matters!)

We also wanted to leave space for future expansion: