🔬 Deep Research Skill

Browse hundreds of sources in parallel — like DeepSeek, but inside your AI coding agent. No API keys, no limits, completely free.

---

⭐️ Support This Project ⭐️

This is a free, open-source skill built by a solo founder. If it saves you time or money, please:

Every star & sponsor helps a solo founder keep building free tools for everyone. 🙏

---

Like DeepSeek's web research, but inside your AI coding agent. This skill gives your AI agent the power to search, fetch, and synthesize information from hundreds of web pages in parallel — completely free, no API keys, no accounts, no limits. One prompt → agent autonomously searches, reads 100+ sources, and returns a synthesized report with citations.

🔥 Why This Exists

Firecrawl is great, but:

• ❌ Needs API key + account
• ❌ Free tier limited (500K credits)
• ❌ Self-host requires Docker + infrastructure
• ❌ AGPL license

Deep Research Skill uses what OpenCode already has built-in — websearch (powered by Exa, free no API key) and webfetch (built-in, free) — and orchestrates them into a massive parallel research pipeline.

✨ Features

Feature	Description
🔍 Multi-Query Search	Searches from multiple angles for comprehensive coverage
⚡ Parallel Fetching	Fetches 100+ pages simultaneously like DeepSeek
🧠 AI-Powered Synthesis	Agent reads & synthesizes all content intelligently
📊 Structured Output	Comparison tables, deep dives, summaries with citations
🔓 Completely Free	No API keys, no accounts, no credit limits
📚 Research Templates	Comparison, technical, market research — ready to use
🤖 Autonomous Workflow	One prompt → search → fetch → synthesize → report

🚀 Installation

One-Liner (All Agents — Recommended)

curl -fsSL https://raw.githubusercontent.com/FMATheNomad/deep-research-skill/main/skills/deep-research/SKILL.md \
  -o ~/.agents/skills/deep-research/SKILL.md

~/.agents/skills/ is compatible with OpenCode, Claude Code, Cursor, Codex, Windsurf, Cline, and all tools that support the Agent Skills standard.

OpenCode

mkdir -p ~/.config/opencode/skills/deep-research
curl -fsSL https://raw.githubusercontent.com/FMATheNomad/deep-research-skill/main/skills/deep-research/SKILL.md \
  -o ~/.config/opencode/skills/deep-research/SKILL.md

Claude Code

mkdir -p ~/.claude/skills/deep-research
curl -fsSL https://raw.githubusercontent.com/FMATheNomad/deep-research-skill/main/skills/deep-research/SKILL.md \
  -o ~/.claude/skills/deep-research/SKILL.md

Prerequisites

• OpenCode installed or any AI coding agent
• Start with OPENCODE_ENABLE_EXA=1 opencode to enable websearch (OpenCode only)

🔬 How It Works

graph TD
    A[User: One Research Prompt] --> B[websearch: 5-10 queries from different angles]
    B --> C[Collect 100+ URLs across all queries]
    C --> D[webfetch: ALL 100+ pages in parallel]
    D --> E[AI agent reads & synthesizes every source]
    E --> F[Structured Report with Citations]
    
    style A fill:#6C47FF,color:#fff
    style B fill:#4CAF50,color:#fff
    style D fill:#FF9800,color:#fff
    style F fill:#f44336,color:#fff

Loading

Example Session (Real — 100+ Sources)

You: "Deep research about latest DNA sequencing technology, search 50+ sources"

Agent (deep research activated):
  → websearch("DNA sequencing 2025 2026 breakthroughs")
  → websearch("next generation sequencing technology latest")
  → websearch("long read sequencing pacbio nanopore")
  → websearch("DNA sequencing market 2026")
  → websearch("AI in genomics 2025")
  → Collected 120+ URLs from Wikipedia, Nature, Illumina, PacBio, etc.

  → Spawning 3 sub-agents in parallel:
     → Agent 1: Research Illumina + Element Biosciences + MGI
     → Agent 2: Research Oxford Nanopore + PacBio HiFi
     → Agent 3: Research AI applications + market trends
     
  → All 3 agents fetching 40+ pages each in parallel...
  
  → Synthesizing all findings from 100+ sources...
  
  → 📋 Research: DNA Sequencing Technology 2025-2026
  
    Summary: 3 technology generations coexisting...
    
    Key Findings:
    1. Illumina 5-base solution (2026) — detect 5 DNA bases in 1 assay
    2. PacBio HiFi — 99.9% accuracy, $345/genome
    3. Oxford Nanopore — portable, real-time, ultra-long reads
    4. Element AVITI — new chemistry disrupting pricing
    
    Sources (100+):
    - Wikipedia — history & methods
    - Illumina — NGS technology
    - Oxford Nanopore — nanopore tech
    - PacBio — HiFi sequencing
    - Nature — cancer research papers
    - Google Scholar — 59+ cited papers
    - Element Biosciences — Avidite chemistry

📋 Research Templates

Comparison Research

"Compare PostgreSQL and Supabase for production use"
→ Searches: "PostgreSQL vs Supabase", "Supabase review", "PostgreSQL production tips"
→ Fetches 15+ comparison articles, reviews, docs
→ Returns comparison table: features, pricing, pros/cons, use cases

Technical Deep Dive

"How does WebRTC work under the hood?"
→ Searches: "WebRTC explained", "WebRTC architecture", "WebRTC STUN TURN"
→ Fetches 20+ technical articles, RFCs, implementation guides
→ Returns comprehensive technical explanation with diagrams

Market Research

"Analyze the AI code generation market in 2026"
→ Searches: "AI code generation market size", "AI coding tools comparison", "GitHub Copilot vs alternatives"
→ Fetches 25+ market reports, reviews, funding news
→ Returns market analysis with trends, competitors, opportunities

⚡ Pro Tips

1. Enable websearch — start OpenCode with OPENCODE_ENABLE_EXA=1 opencode
2. Ask for 100+ sources — "deep research about X, search 100+ sources"
3. Use sub-agents — "use 3 sub-agents, each researching a different aspect"
4. Specify format — "compare X and Y in a table", "give me a structured report"
5. Iterate — ask follow-up questions to go deeper
6. Combine with agent-browser — for pages that need JS/clicking, use agent-browser then webfetch

🆚 DeepSeek vs Firecrawl

Feature	DeepSeek App	Firecrawl	Deep Research Skill (Ours)
💰 Cost	Free (limited)	Free tier (500K credits)	Unlimited free
🔑 API Key	Not needed	Required	Not needed
🌐 Hundreds of sources	✅ Yes	✅ Yes	✅ Yes — agent batches
🧠 AI synthesis	✅ Built-in	Separate API	✅ Built into agent
🔌 Works inside coding agent	❌ No	❌ No (separate CLI)	✅ Native OpenCode/Claude
🔍 Search	Built-in	Built-in	via websearch (Exa)
📄 Scrape	Built-in	Built-in	via webfetch
📜 License	Proprietary	AGPL-3.0	MIT
🛠 Customizable	❌ No	Limited	✅ Full control

Bottom line: DeepSeek is great for chat-based research. Firecrawl is better for automated crawling. Deep Research Skill is the only one that gives you DeepSeek-like massive parallel research directly inside your AI coding agent — completely free and open source.

🗺 Roadmap

• Bulk mode — research 10+ topics in one session
• PDF support — fetch and parse academic PDFs
• Citation export — BibTeX, APA, MLA format
• Research report generation — export to markdown/PDF
• Scheduled research — monitor topics over time
• Multi-language — research in Indonesian, Japanese, etc.

📁 Project Structure

deep-research-skill/
├── .github/
│   ├── workflows/validate.yml
│   └── FUNDING.yml
├── skills/
│   └── deep-research/
│       └── SKILL.md
├── README.md
├── LICENSE
└── .gitignore

🤝 Contributing

Add new research templates, improve the pipeline, or suggest features. See CONTRIBUTING.md.

---

⭐️ Support the Project ⭐️

Built by a solo founder who got tired of API keys and credit limits.
If this skill helps you, please support it — every bit counts:

---

Free research for everyone. Open source. MIT licensed.