On this page
Research is one of the most time-consuming tasks in any knowledge work. What takes a human researcher 10+ hours can now be done in 10 minutes with CrawlForge's deep_research tool. This guide shows you how.
The Research Problem
Manual research is brutal:
| Task | Manual Time | Manual Steps |
|---|---|---|
| Topic research | 4-8 hours | Search, read, note-take, verify, synthesize |
| Market analysis | 6-12 hours | Find sources, extract data, compare, analyze |
| Due diligence | 10-20 hours | Company research, news, financials, verify |
| Literature review | 20-40 hours | Find papers, read, cite, synthesize |
The pattern is always the same:
- Search for relevant sources
- Read and extract key information
- Verify across multiple sources
- Detect conflicting information
- Synthesize into actionable insights
Each step is tedious. Each step is automatable.
The Deep Research Solution
CrawlForge's deep_research tool handles the entire research pipeline:
// One command to research any topic:
await deepResearch({
topic: "Current state of quantum computing commercialization",
maxDepth: 5,
maxUrls: 50,
enableSourceVerification: true,
enableConflictDetection: true,
enableSynthesis: true,
credibilityThreshold: 0.3
});What happens behind the scenes:
- Query Expansion - Generates related search queries
- Multi-Source Search - Searches across Google, news, and academic sources
- Content Extraction - Scrapes and cleans relevant pages
- Source Verification - Scores credibility of each source
- Conflict Detection - Identifies disagreements between sources
- Synthesis - Generates comprehensive summary with citations
Real-World Example: Market Research
Let's walk through a real research task.
The Request
Research the web scraping tools market in 2026:
- Market size and growth
- Key players and market share
- Pricing trends
- Technology trends
- Predictions for 2026
Manual Approach (Estimated: 8 hours)
- Search Google for "web scraping market size 2026" (30 min)
- Find Statista, Gartner, or similar reports (30 min)
- Search for competitor information (1 hour)
- Visit each competitor's website (2 hours)
- Extract pricing from each (1 hour)
- Search for technology trends (1 hour)
- Verify information across sources (1 hour)
- Synthesize into report (1 hour)
Total: ~8 hours
CrawlForge Approach (Actual: 8 minutes)
// In Claude Code:
"Use deep_research to analyze the web scraping tools market in 2026.
Include market size, key players, pricing trends, and technology trends.
Verify all claims across multiple sources and flag any conflicts."
// Configuration used:
{
"topic": "web scraping tools market analysis 2026",
"maxUrls": 50,
"researchApproach": "broad",
"enableSourceVerification": true,
"enableConflictDetection": true,
"outputFormat": "comprehensive"
}Actual time: 8 minutes, 23 seconds Credits used: 10
The Output
# Web Scraping Tools Market Analysis 2026
## Executive Summary
The web scraping tools market reached $X billion in 2026, growing at
Y% CAGR. Key trends include AI integration, MCP protocol adoption,
and shift toward managed services.
## Market Size & Growth
- 2026 Market Size: $X billion [Source: Statista, Grand View Research]
- CAGR 2026-2030: Y% [Source: Markets and Markets]
- ⚠️ CONFLICT: Gartner estimates X+0.5B, while Statista estimates X-0.3B
## Key Players
| Company | Est. Market Share | Key Differentiator |
|---------|-------------------|-------------------|
| Bright Data | 25-30% | Largest proxy network |
| Apify | 15-20% | Actor marketplace |
| ScraperAPI | 10-15% | Simplicity |
| Firecrawl | 5-10% | AI focus |
| CrawlForge | New entrant | MCP native, 20 tools |
## Pricing Trends
- Average price per 1000 requests: $X-Y
- Trend: Moving toward credit-based pricing
- Enterprise deals increasingly common
## Technology Trends
1. AI/LLM Integration (mentioned in 23/50 sources)
2. MCP Protocol Adoption (mentioned in 12/50 sources)
3. Anti-bot evasion (mentioned in 34/50 sources)
4. Headless browser optimization (mentioned in 18/50 sources)
## 2026 Predictions
- Market expected to reach $X billion
- Consolidation expected among mid-tier players
- AI-native tools to capture 30%+ market share
---
## Source Credibility Scores
| Source | Credibility | Type |
|--------|-------------|------|
| Statista | 0.9 | Data Provider |
| Gartner | 0.95 | Analyst |
| TechCrunch | 0.7 | News |
| Company Websites | 0.6 | Commercial |
## Detected Conflicts
1. Market size estimates vary by ±$500M between Gartner and Statista
2. Growth rate estimates: Optimistic (15%) vs Conservative (8%)
## All Sources (50)
1. https://statista.com/...
2. https://gartner.com/...
[...]ROI Calculation
| Metric | Manual | CrawlForge |
|---|---|---|
| Time | 8 hours | 8 minutes |
| Cost (at $50/hr) | $400 | $0.10 (10 credits) |
| Sources checked | 10-15 | 50 |
| Conflict detection | Manual | Automatic |
| Citations | Manual | Automatic |
Time savings: 60x Cost savings: 4,000x
Configuration Options
Research Approaches
// Broad research (default) - good for market analysis
{ "researchApproach": "broad" }
// Focused research - good for specific questions
{ "researchApproach": "focused" }
// Academic research - prioritizes scholarly sources
{ "researchApproach": "academic" }
// Current events - prioritizes recent news
{ "researchApproach": "current_events" }
// Comparative - good for X vs Y analysis
{ "researchApproach": "comparative" }Source Type Filtering
{
"sourceTypes": ["academic", "news", "government"]
// Options: academic, news, government, commercial, blog, wiki, any
}Credibility Threshold
{
"credibilityThreshold": 0.5 // Only include sources scoring > 0.5
// Range: 0.0 (all sources) to 1.0 (only highest credibility)
}Output Formats
// Full report with all details
{ "outputFormat": "comprehensive" }
// Quick summary only
{ "outputFormat": "summary" }
// Just the sources and citations
{ "outputFormat": "citations_only" }
// Focus on disagreements between sources
{ "outputFormat": "conflicts_focus" }Use Cases
1. Due Diligence
"Research [Company Name] for due diligence:
- Company history and founding team
- Funding history and investors
- Product/service overview
- Recent news and press
- Customer reviews and complaints
- Competitors and market position
Flag any red flags or concerning information."2. Competitive Analysis
"Deep research on [Competitor] vs [Our Company]:
- Feature comparison
- Pricing comparison
- Customer sentiment
- Recent developments
- Market positioning"3. Technology Assessment
"Research the current state of [Technology]:
- How it works (technical overview)
- Current adoption levels
- Key players and implementations
- Limitations and challenges
- Future outlook"4. Investment Research
"Research [Stock/Crypto] for investment:
- Fundamental analysis
- Recent news and developments
- Analyst opinions (aggregate bull/bear cases)
- Risk factors
- ⚠️ Important: Flag conflicting analyst opinions"5. Academic Literature Review
"Literature review on [Topic]:
- Use academic source filter
- Key papers and authors
- Main findings and consensus
- Open questions and debates
- Recent developments (past 2 years)"
{
"researchApproach": "academic",
"sourceTypes": ["academic"],
"credibilityThreshold": 0.7
}Conflict Detection Deep Dive
One of deep_research's most valuable features is automatic conflict detection:
{
"enableConflictDetection": true
}What It Detects
| Conflict Type | Example |
|---|---|
| Numerical disagreements | "Market size $5B" vs "$7B" |
| Date discrepancies | "Founded 2020" vs "Founded 2019" |
| Factual contradictions | "Supports X" vs "Does not support X" |
| Opinion divergence | "Will succeed" vs "Will fail" |
How It Works
- Extracts claims from each source
- Normalizes claim formats
- Compares across sources
- Flags disagreements
- Shows source for each position
Example Output
## Detected Conflicts
### Conflict 1: Market Size Estimates
- **Position A:** $5.2 billion (Statista, Forbes)
- **Position B:** $7.1 billion (Grand View Research)
- **Resolution:** Difference may be due to market definition scope
### Conflict 2: Company Valuation
- **Position A:** Valued at $500M (TechCrunch, 2024)
- **Position B:** Valued at $750M (Company press release, 2026)
- **Resolution:** Valuation increased between reportsBest Practices
1. Be Specific with Topics
// Too broad (may return unfocused results):
"Research AI"
// Better:
"Research the current state of AI code generation tools,
focusing on GitHub Copilot, Cursor, and Claude Code"2. Set Appropriate Depth
// Quick overview (faster, cheaper):
{ "maxDepth": 2, "maxUrls": 20 } // ~5 credits worth
// Comprehensive research (thorough):
{ "maxDepth": 5, "maxUrls": 50 } // 10 credits
// Exhaustive (for critical decisions):
{ "maxDepth": 10, "maxUrls": 100 } // May exceed 10 credits3. Filter by Recency
{
"includeRecentOnly": true // Focus on recent sources
}4. Verify Critical Information
For high-stakes decisions, always verify critical claims:
"Verify the claim that [specific claim] by checking:
1. Primary source (if cited)
2. At least 2 independent sources
3. Any official documentation"Combining with Other Tools
Deep research works best as part of a workflow:
// 1. Deep research for initial findings (10 credits)
"Deep research on [topic]"
// 2. Follow up with specific URL scraping (1-2 credits each)
"Fetch the full report from [url mentioned in research]"
// 3. Monitor for updates (3 credits)
"Track changes on [source url] and alert me to updates"Limitations
Be aware of what deep_research can't do:
| Limitation | Workaround |
|---|---|
| Can't access paywalled content | Use direct URLs if you have access |
| Real-time data (stocks, etc.) | Use specialized APIs |
| Very recent events (< 1 hour) | Use news APIs |
| Private company data | Combine with official filings |
| Subjective judgments | Use as input for human decision |
Getting Started
Ready to try deep research? Here's the fastest path:
# 1. Sign up for free (1,000 credits)
# Visit: https://crawlforge.dev/signup
# 2. Set up CrawlForge MCP
npm install -g crawlforge-mcp-server
npx crawlforge-setup
# 3. In Claude Code, try:
"Deep research on [your topic of interest]"Your free tier includes 100 deep research queries (10 credits each).
Related Resources: