google-researcher-mcp

# Google Researcher MCP - Full LLM Context

> Comprehensive reference for AI assistants using the Google Researcher MCP server.
> This file contains everything needed to effectively use this MCP for research tasks.

---

## Server Overview

The Google Researcher MCP server enables AI assistants to:
- Search the web via Google Custom Search API
- Read and extract content from any URL (including JavaScript-rendered pages)
- Extract YouTube video transcripts
- Parse documents (PDF, DOCX, PPTX)
- Search academic papers with citations
- Search patents for IP research
- Track multi-step research workflows

---

## Tool Reference

### search_and_scrape (RECOMMENDED)

**Use this for most research tasks.** Combines search and content retrieval in one call.

```json
{
  "name": "search_and_scrape",
  "arguments": {
    "query": "climate change effects 2024",
    "num_results": 5,
    "deduplicate": true
  }
}
```

| Parameter | Type | Default | Description |
|-----------|------|---------|-------------|
| query | string | required | Search query (1-500 chars) |
| num_results | number | 3 | Sources to fetch (1-10) |
| deduplicate | boolean | true | Remove duplicate content |
| include_sources | boolean | true | Append source URLs |
| max_length_per_source | number | 50KB | Max chars per source |
| total_max_length | number | 300KB | Max total chars |
| filter_by_query | boolean | false | Only paragraphs with query terms |

**Response includes:**
- Combined content from all sources
- Quality scores for each source
- Citations (APA/MLA format)
- Size metadata (estimatedTokens, truncated)

---

### scrape_page

Extract content from a specific URL. Auto-detects content type.

```json
{
  "name": "scrape_page",
  "arguments": {
    "url": "https://example.com/article",
    "mode": "full"
  }
}
```

**Supported content types:**
- Web pages (static HTML and JavaScript SPAs)
- YouTube videos (extracts transcript)
- PDF documents
- DOCX documents
- PPTX presentations

**Special capabilities:**
- Automatically uses browser rendering for JavaScript sites
- Preview mode to check page size before full fetch
- Citation extraction from page metadata

**For Google Patents:**
```json
{
  "name": "scrape_page",
  "arguments": {
    "url": "https://patents.google.com/?assignee=CompanyName"
  }
}
```
This returns the full patent list for a company - more comprehensive than patent_search.

---

### google_search

Returns URLs only (not content). Use when you need to process pages yourself.

```json
{
  "name": "google_search",
  "arguments": {
    "query": "TypeScript best practices",
    "num_results": 5,
    "time_range": "month",
    "site_search": "github.com"
  }
}
```

---

### google_news_search

Search recent news with freshness filtering.

```json
{
  "name": "google_news_search",
  "arguments": {
    "query": "AI regulations",
    "freshness": "week",
    "sort_by": "date",
    "num_results": 5
  }
}
```

| freshness | Description |
|-----------|-------------|
| hour | Last 60 minutes |
| day | Last 24 hours |
| week | Last 7 days |
| month | Last 30 days |
| year | Last 365 days |

---

### google_image_search

Find images with filtering options.

```json
{
  "name": "google_image_search",
  "arguments": {
    "query": "architecture diagram microservices",
    "type": "lineart",
    "size": "large"
  }
}
```

| type | Description |
|------|-------------|
| photo | Photographs |
| lineart | Diagrams, drawings |
| clipart | Clip art |
| face | Face images |
| animated | GIFs |

---

### academic_search

Search academic papers with citations.

```json
{
  "name": "academic_search",
  "arguments": {
    "query": "transformer neural networks",
    "num_results": 5,
    "year_from": 2020,
    "source": "arxiv"
  }
}
```

**Sources:** all, arxiv, pubmed, ieee, nature, springer

**Returns:**
- Paper titles, authors, abstracts
- Pre-formatted citations (APA, MLA, BibTeX)
- PDF links when available

---

### patent_search

Search Google Patents via Custom Search API.

```json
{
  "name": "patent_search",
  "arguments": {
    "query": "video streaming adaptive bitrate",
    "search_type": "prior_art",
    "num_results": 5
  }
}
```

**For comprehensive company patent research:** Use `scrape_page` on `patents.google.com/?assignee=CompanyName` instead - it returns more complete results.

---

### sequential_search

Track multi-step research across multiple API calls.

```json
{
  "name": "sequential_search",
  "arguments": {
    "searchStep": "Starting research on quantum computing",
    "stepNumber": 1,
    "totalStepsEstimate": 5,
    "nextStepNeeded": true
  }
}
```

Use for complex investigations requiring 3+ searches with different angles.

---

## Prompts Reference

### Basic Prompts

#### comprehensive-research
Multi-source research with synthesis and citations.
```json
{ "name": "comprehensive-research", "arguments": { "topic": "climate change solutions", "depth": "deep" } }
```
depth: quick (3 sources), standard (5), deep (8)

#### fact-check
Verify claims against authoritative sources.
```json
{ "name": "fact-check", "arguments": { "claim": "Coffee is the most traded commodity", "sources": 4 } }
```

#### summarize-url
Summarize content from a URL.
```json
{ "name": "summarize-url", "arguments": { "url": "https://example.com", "format": "bullets" } }
```
format: brief, detailed, bullets

#### news-briefing
Current news summary on a topic.
```json
{ "name": "news-briefing", "arguments": { "topic": "AI regulations", "timeRange": "week" } }
```

### Advanced Prompts

#### patent-portfolio-analysis
Analyze company patent holdings including subsidiaries.
```json
{ "name": "patent-portfolio-analysis", "arguments": { "company": "Kaltura", "includeSubsidiaries": true } }
```

#### competitive-analysis
Compare multiple companies/products.
```json
{ "name": "competitive-analysis", "arguments": { "entities": "React, Vue, Angular", "aspects": "performance, ecosystem" } }
```

#### literature-review
Academic literature synthesis with proper citations.
```json
{ "name": "literature-review", "arguments": { "topic": "machine learning in healthcare", "yearFrom": 2020, "sources": 8 } }
```

#### technical-deep-dive
In-depth technical investigation.
```json
{ "name": "technical-deep-dive", "arguments": { "technology": "WebAssembly", "focusArea": "architecture" } }
```
focusArea: architecture, implementation, comparison, best-practices, troubleshooting

---

## Resources (via resources/read)

| URI | Description |
|-----|-------------|
| search://recent | Last 20 search queries with timestamps |
| search://session/current | Active sequential search session state |
| stats://cache | Cache hit rates and statistics |
| stats://events | Event store statistics |
| stats://resources | Resource cache statistics |
| config://server | Server capabilities and configuration |

---

## Quality Scoring

Sources from search_and_scrape are scored by:

| Factor | Weight | Description |
|--------|--------|-------------|
| Relevance | 35% | Query term matching |
| Freshness | 20% | Publication recency |
| Authority | 25% | Domain reputation (.gov, .edu, known sources) |
| Content Quality | 20% | Length, structure, readability |

---

## Key Behaviors

### Caching
- Search results: 30 minutes
- Scraped pages: 1 hour
- Repeated queries are fast and don't count against API limits

### SPA Rendering
Automatically uses browser rendering for JavaScript-heavy sites:
- patents.google.com
- scholar.google.com
- twitter.com / x.com
- linkedin.com

### Graceful Failures
If some sources fail, you still get results from successful ones.

### Citations
All scraped content includes citation metadata in APA/MLA format.

---

## Best Practices

1. **Start with search_and_scrape** for most research tasks
2. **Use preview mode** before fetching large pages
3. **For patents**: Use scrape_page on patents.google.com/?assignee=X
4. **Cite sources**: Citations are automatically provided
5. **Check quality scores**: Higher scores = more reliable sources
6. **Use prompts** for structured research workflows

---

## Example Workflows

### Research Workflow
1. Use `comprehensive-research` prompt for topic overview
2. Follow up with `academic_search` for peer-reviewed sources
3. Use `scrape_page` on key URLs for full content

### Patent Analysis Workflow
1. Use `patent-portfolio-analysis` prompt
2. Scrape patents.google.com/?assignee=CompanyName for full list
3. Search for subsidiaries with search_and_scrape
4. Compile comprehensive patent table

### Competitive Analysis Workflow
1. Use `competitive-analysis` prompt
2. Follow up with `technical-deep-dive` for specific technologies
3. Use `google_news_search` for recent developments

---

## Limits

| Limit | Value |
|-------|-------|
| Max scrape content | 50 KB |
| Max combined research | 300 KB |
| Max document size | 10 MB |
| Max search results | 10 per query |
| Google API rate | 100 queries/day |