A content research agent that discovers, catalogs, and promotes public content about your product, technology, open-source project, or tool across the developer ecosystem. Scans 14+ public sources, filters for quality, generates reports with topic tags, and drafts ready-to-post social media content — all configured through a single onboarding conversation.
Run it interactively in your editor (VS Code, Claude Code, Cursor, and more) or from a browser with the built-in Web UI. The agent handles all standard content work; the web UI adds dashboards, bulk operations, and visual triage. See docs/SURFACES.md for which surface to use when.
git clone https://github.com/jaydestro/content-scout.git
cd content-scout
Then pick a path:
Requires Node.js 22.13 or later.
cd tools/web-ui
npm install
npm start
Open http://localhost:4477. On first start, Content Scout creates a local
SQLite database at .local/state/content-scout.db and imports any existing
reports, social drafts, and configs automatically. No database server or
credentials are required. The onboarding wizard then saves your config and
lets you launch scans, view reports, and manage .env keys. See
tools/web-ui/README.md for the full reference.
Open the repo in your AI chat tool and run /scout-onboard (VS Code) or say "scout onboard" (Claude Code, Cursor, Windsurf, Cline, Copilot terminal chat). See docs/EDITORS.md for per-tool startup, slash-command list, and troubleshooting.
| Command | What it does |
|---|---|
/scout-onboard |
Set up the agent for a new product, technology, or project |
/scout-scan |
Scan for content — specify a topic slug or scan all |
/scout-post |
Generate social posts from a URL or report item |
/scout-calendar |
Create a weekly posting schedule |
Plus /scout-creators, /scout-doctor, /scout-keys, /scout-seo, /scout-originality, /scout-reddit-import, /scout-alt, /scout-vision — see docs/EDITORS.md.
14 standard sources plus optional custom sources (vendor blogs, update feeds, docs, influencer blogs):
- No auth needed: Dev.to, Medium, Hashnode, DZone, C# Corner, InfoQ, GitHub, Stack Overflow, Hacker News, LinkedIn
- Free auth: YouTube (API key), Reddit (RSS by default; OAuth optional), Bluesky (app password)
- Paid auth (optional): X/Twitter ($200/mo Basic plan)
- Logged-in browser (recommended for X / LinkedIn / Reddit): tools/browser-scan/ attaches to your real browser (Edge / Chrome / Brave / Vivaldi / Arc / Opera — auto-detects your OS default) over the Chrome DevTools Protocol, so logged-in scrapes work with no anti-bot trips. Either click 🌐 Browser scan (Layer 0) in the web UI's Run view, or run
node tools/browser-scan/launch-edge.mjsonce andnode tools/browser-scan/index.mjs scan --slug {slug}before eachscout scan. The resulting JSON sidecars become Layer 0 for those three platforms.
All API keys are optional — without them, the agent skips those sources. Keys live in .env (not in config files). See docs/API-KEYS.md and docs/SOURCES.md.
Content Scout vendors the humanizer skill (MIT) at .claude/skills/humanizer/SKILL.md and runs every generated social post through it before saving. The skill strips the common tells of AI-generated copy — promotional adjectives, AI-vocabulary words ("delve", "underscore", "showcase"), significance inflation, em-dash overuse, negative parallelisms, and chatbot openers like "Excited to share…". Posts read like a practitioner wrote them, not like an LLM autocompleted a marketing brief.
No setup is required — the skill is part of the repo and loads automatically with /scout-post and /scout-scan. If your editor lists user-level skills separately and prefers them over repo-vendored ones, install the upstream skill from https://github.com/blader/humanizer and the same patterns will apply.
The same signals that power the humanizer drive an originality review. Turn on Originality scoring in your config and /scout-scan adds a transparent 0-10 read of how human-written vs. AI-generated each text-bearing item looks (never a verdict — it never drops or down-ranks an item). For a one-off check, run /scout-originality (or the web UI Tools → Originality tab) on any URL: it scores the AI-writing signals and runs a verbatim-overlap check against the official docs so you can tell original analysis from repackaged documentation.
Content Scout can roll every in-month content report into a single monthly roundup — one regenerable index per calendar month (official posts, videos, articles, and sample repos), deduped and with author bylines filled in. Generate it from the web UI's Reports → Roundup tab; it saves to reports/{YYYY-MM}-{slug}-roundup.md and overwrites in place each time you regenerate.
Pick one role, multiple roles (merged), or a custom role during onboarding. Each role enables different features and report sections — Program Manager, Product Manager, Social Media Manager, Product Marketer, Developer Advocate, Community Manager, Technical Writer. See docs/WORKFLOW.md for the full feature matrix.
The examples/ folder contains sample outputs using Azure Cosmos DB:
- example-config.md — completed product configuration
- example-report.md — monthly content report
- example-social-posts.md — LinkedIn and X posts
- example-posting-calendar.md — 2-week posting schedule
| Doc | What's in it |
|---|---|
| Workflow | End-to-end guide: onboarding, scanning, posting, analysis |
| Editors | Per-tool setup, slash commands, troubleshooting |
| Content Sources | All 14 standard sources, custom sources, scanning order |
| API Keys | YouTube, Reddit, Bluesky, X/Twitter setup |
| Architecture | Subagents, quality filters, thumbnails, file structure |
See CONTRIBUTING.md.
MIT