Learn a skill

Deep Web Research

Teach your agent to read the open web (sites, Reddit, YouTube, GitHub) and build cited research notebooks.

What we found works best

Use Firecrawl MCP for reliable page and site extraction. Add Agent-Reach for social platforms (read-only), and Open Notebook to keep sources organized.

Firecrawl handles messy sites well (it's paid, so budget for it). Agent-Reach covers platforms that ordinary fetch can't. Open Notebook keeps your research private and self-hosted.

Options we compared

OptionCost / accountCan act?Source
✅ Firecrawl MCP
Scrape, crawl and extract to clean markdown
Firecrawl credits
Firecrawl API key
spends creditsfirecrawl/firecrawl-mcp-server
MIT · ★ 7.6k · pushed 2026-10-14
Agent-Reach
Read and search X, Reddit, YouTube and GitHub from one CLI
free per README (unverified)
varies per platform
read-only use only: no postingpanniantong/agent-reach
MIT · ★ 97k · pushed 2026-10-08
Open Notebook
A self-hosted NotebookLM-style research notebook
free (self-host)
LLM keys
nonelfnovo/open-notebook
MIT · ★ 40k · pushed 2026-10-13

Steps

  1. Install the research-agent blueprint first (it covers the citation rules).
  2. Connect Firecrawl with a monthly credit cap.
  3. Optional: Agent-Reach for social reading, in read-only mode.
  4. Research one question into a cited brief.
  5. Archive the sources in Open Notebook.

Traps to avoid

Safety

🔒 Pro plan: guided install

Exact commands for Claude Code, Codex, Grok Bot and Gemini CLI, pinned commits, one step at a time with next_lesson and check_lesson, plus updates.

Stars and last push were checked against the GitHub API. Install commands come from vendor and README docs and have not been executed by us. Marked "unverified" where we couldn't confirm.

Ask the teacher for a plan for your agent →