/projects/competitorresearch
A single-task competitive intelligence agent: enter a competitor URL, scrape key pages with Firecrawl, extract structured company context, then ask a ToolLoopAgent for positioning, pricing, features, tech stack, weaknesses and battlecards.
Competitive research is usually manual, stale, and hard to compare across companies. Teams need structured intelligence from competitor sites, not screenshots or generic summaries.
The Competitor Research Agent maps and scrapes key pages, extracts normalized company context, then runs analysis tools for positioning, pricing, features, tech stack, weaknesses, and battlecards.
The difficult pieces are resilient scraping, schema-normalized extraction, public rate limiting, inconsistent pricing pages, and producing analysis that sales or product teams can actually use.
The workflow combines Firecrawl mapping, OpenAI structured extraction, and specialized analysis tools so each output is tied to a tactical competitive-intelligence job.
This shows Stefan's strengths in market-intelligence automation, scraping architecture, structured extraction, battlecard workflows, pricing analysis, and AI product tooling for GTM teams.
The Competitor Research Agent turns public websites into structured market intelligence. Its value is in separating scraping, normalization, analysis, and battlecard synthesis into distinct tactical jobs.
Competitive intelligence starts with finding the right pages. Mapping pricing, features, about, proof, and comparison pages gives the model better source material than a single homepage scrape.
A normalized competitor profile gives downstream analysis a stable base. Without that schema, pricing, positioning, feature, and weakness analysis becomes inconsistent across companies.
Separate tools for positioning, pricing, features, tech stack, weaknesses, and battlecards match real GTM workflows. Each tool answers a different business question.
The final output helps founders and sales teams understand where to compete, how to handle objections, and which gaps are product opportunities rather than trivia.
Works well for a single-task research app because scraping, extraction, and chat can share one project boundary.
Supports rich views for pricing, features, stack badges, weaknesses, and battlecards.
Keeps the scraped profile, tool calls, and UI view models aligned as the analysis expands.
Runs the follow-up analyst loop with tool calling, streaming, and model-provider flexibility.
Handles the conversational competitor analyst path through a fast tool-loop model.
Used for the structured extraction pass where a clean CompetitorData object matters.
Chosen because competitor intelligence starts with mapping and scraping real public pages.
Adds a small, durable rate-limit guard for public scraping and agent calls.
Defines the competitor schema so extraction errors do not leak into the UI.
Coordinates URL validation, rate limiting, Firecrawl mapping, page scraping, and extraction.
Finds the pricing, features, about, and proof pages that usually carry market intelligence.
Converts scraped markdown into the normalized competitor profile used by the app.
Extracts value propositions, audience, messaging themes, and category position.
Builds a structured view of tiers, billing model, trials, and free-plan details.
Turns page content into feature groups, gaps, and comparison-ready matrices.
Detects frameworks, infrastructure, analytics, auth, payments, and overall modernity.
Highlights exploitable gaps, complaints, and positioning opportunities.
Creates objections, responses, winning tactics, and traps to avoid.