The Firecrawl team's social media surveillance drone

What it does Trend Finder is a cron-driven TypeScript application that monitors specific Twitter/X accounts and websites for new posts or releases. It pipes that content through an LLM—Together AI by default, though DeepSeek and OpenAI are also supported—to flag emerging trends, product launches, or shifts in conversation. If the model judges something significant, it fires a notification into a Slack or Discord channel with context and links. It was built by the Firecrawl marketing team to automate the manual slog of social listening.

The interesting bit Instead of tracking hashtag velocity or engagement spikes, the tool outsources all trend judgment to an LLM, feeding it scraped posts and pages to decide what merits an alert. That makes the signal-to-noise ratio entirely a function of the model’s reasoning and the instructions it receives—which the README never details, leaving the actual detection logic a black box.

Key highlights

• Monitors both Twitter/X accounts (via the X API) and websites (via Firecrawl’s /extract) on a cron schedule.
• Supports multiple LLM backends: Together AI, DeepSeek, and OpenAI.
• Sends formatted alerts to Slack or Discord with source links and context.
• Built in TypeScript with a minimal controllers/services structure and Docker support.
• Explicitly warns that the free X API tier limits you to checking one account every fifteen minutes.

Caveats

• The free X API tier is severely rate-limited, making the social media half of the tool impractical without a paid plan.
• The README describes the AI analysis step in broad strokes—“identifies emerging trends”—without explaining the actual detection logic, thresholds, or how it avoids false positives.
• It requires a constellation of external API keys and webhooks, and the README never mentions local state, caching, or a database.

Verdict Marketing teams and developer advocates who want an LLM-curated digest of influencer chatter will find this handy—assuming they already pay for the X API and enjoy herding a handful of API keys. If you need transparent scoring, historical data, or rigorous false-positive controls, look elsewhere.