GuideSlate Editorial Library — Full English AI tool tutorials are 100% free · Companion .zip starter kits $9.90 one-time All kits
n8n & automation Updated Nov 2026 · v3.1 · Saves 6 hrs · Beginner-friendly

Monitor X and Reddit buying-intent threads every 15 minutes with n8n

Skip $99/mo social listening SaaS tools. Poll targeted RSS and search endpoints, score high-intent leads with a low-cost Flash LLM, and push ready-to-reply alerts to Slack.

ARCHITECTURE PIPELINE
01 · Cron 15m Search feeds 02 · Dedup SQLite hash 03 · Flash LLM Score >= 8/10 04 · Webhook Slack alert

Why default setups fall short

Most social keyword trackers flood your inbox with self-promotional noise. By chaining an n8n schedule node with lightweight search feeds and a strict JSON intent classifier running on DeepSeek Flash or Gemini Flash-Lite, you pay less than $0.40 a month in tokens and only get pinged when a real human asks for a tool recommendation or complains about a competitor.

Prerequisites before starting

  • Self-hosted n8n (Docker) or n8n Cloud starter
  • An OpenAI-compatible or DeepSeek API key ($1 credit lasts months)
  • A Slack incoming webhook URL or Telegram bot token

Step-by-step implementation

STEP 01

1. Target pain-first query operators instead of broad keywords

Broad keywords like 'CRM' or 'video editor' pull 95% marketing spam. Combine frustration verbs and switching phrases so your feed starts clean before an LLM ever touches it.

("alternative to" OR "migrating from" OR "any tool that" OR "tired of paying")
AND ("invoice" OR "onboarding" OR "workflow") -filter:retweets
STEP 02

2. Deduplicate post IDs locally before calling the model

Use n8n's built-in workflow static data or a lightweight SQLite node to store the last 500 seen post URLs. Skipping already-scored posts cuts token spend by roughly 85%.

const seen = $getWorkflowStaticData('global');
seen.ids = seen.ids || [];
const fresh = items.filter(i => !seen.ids.includes(i.json.guid));
seen.ids = [...fresh.map(i => i.json.guid), ...seen.ids].slice(0, 500);
return fresh;
STEP 03

3. Score buying intent from 1 to 10 with thinking disabled

Call a fast non-reasoning endpoint with temperature 0.1 and force a compact JSON response containing intent_score, buyer_pain, and a non-salesy reply draft.

{
  "model": "deepseek-chat",
  "temperature": 0.1,
  "response_format": { "type": "json_object" }
}
STEP 04

4. Route scores >= 8 to Slack with a one-click open button

Drop anything below 8 silently. Push high-intent matches into a dedicated #buying-signals channel with the author's exact quote and a suggested helpful reply.

More practical field guides

View all 6 guides
Claude & Cursor skills Featured kit
Saves 4.5 hrs · v2.4
Replace generic AI interfaces with a custom SKILL.md design token system

How top solo builders force Claude Code, Cursor, and Antigravity to output studio-grade editorial interfaces instead of default blue-gray Tailwind cards.

01 · Input SKILL.md rules 02 · Tokens 6-hex palette 03 · Guardrail Anti-cliche lint 04 · Output Editorial SSR UI
Companion .zip starter kit (4 files) $9.90 one-time
01-editorial-frontend-SKILL.md (6.4 KB)02-cursor-anti-slop-rules.mdc (3.8 KB)03-architectural-bento-tokens.css (5.1 KB)README-quickstart.md (2.2 KB)
Read free guide
MCP & browser agents
Saves 3.5 hrs · v1.9
The safe 5-server MCP config for browser QA, read-only SQL, and Git

Wire Chrome DevTools, read-only Postgres, scoped filesystem, and GitHub search into Claude Desktop or local coding agents without exposing production write credentials.

01 · Agent Claude / IDE 02 · MCP Router Scoped env 03 · Browser DevTools QA 04 · Database Read-only role
Companion .zip starter kit (4 files) $9.90 one-time
01-claude_desktop_config.json (2.9 KB)02-postgres-readonly-bootstrap.sql (1.6 KB)03-browser-verification-SKILL.md (4.2 KB)README-quickstart.md (1.8 KB)
Read free guide