mcp-sequential-research 1.0.0 → 1.1.1

package/README.md

@@ -229,12 +229,13 @@ This format is designed for downstream claim-mining tools.
 
 Works with other MCP servers:
 
-| Source Type | Recommended MCP |
-|-------------|-----------------|
-| Patents | Google Patents MCP (`search_patents`) |
-| Web | Google Search MCP (Custom Search) |
-| Memory | Memory MCP (`search_nodes`) |
-| Academic | Semantic Scholar API |
+| Source Type | Recommended MCP | Tool |
+|-------------|-----------------|------|
+| Patents | Google Patents MCP | `search_patents` |
+| Web Search | Google Search MCP | `google_search` |
+| Web Scraping | Google Search MCP | `read_webpage` |
+| Memory | Memory MCP | `search_nodes` |
+| Academic | Semantic Scholar API | — |
 
 ## License
 

package/docs/MCP_GUIDANCE.md

@@ -32,14 +32,13 @@ This is the exact **operator loop** Claude Code follows for comprehensive resear
 │     └─→ sequential-research:sequential_research_plan(prompt, constraints) │
 │                        ↓                                              │
 │  2. WEB QUERIES (for each plan.queries where query_family == "web")  │
-│     └─→ google-search:search({query, num: 10})                       │
+│     └─→ google-search:google_search({query, num: 10})                │
 │         Returns: {title, link, snippet}[]                            │
 │                        ↓                                              │
 │  3. SCRAPE WEB CONTENT                                               │
 │     ├─→ Collect top URLs (dedupe)                                    │
-│     ├─→ If multiple: firecrawl:firecrawl_batch_scrape({urls})        │
-│     └─→ If single:   firecrawl:firecrawl_scrape({url})               │
-│         Returns: {markdown, metadata}                                │
+│     └─→ For each URL: google-search:read_webpage({url})              │
+│         Returns: {title, text, url}                                  │
 │                        ↓                                              │
 │  4. PATENT QUERIES (for each plan.queries where query_family == "patent") │
 │     └─→ google-patents:search_patents({query, num_results})          │
@@ -89,7 +88,7 @@ For each query where `query_family == "web"`:
 
 ```json
 {
-  "tool": "google-search:search",
+  "tool": "google-search:google_search",
   "arguments": {
     "query": "photonic computing silicon photonics site:.edu OR filetype:pdf",
     "num": 10
@@ -110,66 +109,69 @@ For each query where `query_family == "web"`:
 }
 ```
 
-### Step 3: Scrape Web Content via Firecrawl MCP
+### Step 3: Scrape Web Content via Google Search MCP `read_webpage`
 
-After collecting search results, extract full content using Firecrawl.
+After collecting search results, extract full content using `read_webpage`.
 
-**For multiple URLs (recommended for efficiency):**
+**For each URL:**
 ```json
 {
-  "tool": "firecrawl:firecrawl_batch_scrape",
+  "tool": "google-search:read_webpage",
   "arguments": {
-    "urls": [
-      "https://example.mit.edu/photonics.pdf",
-      "https://lightmatter.co/technology",
-      "https://ieee.org/article/photonic-computing"
-    ],
-    "options": {
-      "formats": ["markdown"],
-      "onlyMainContent": true
-    }
+    "url": "https://example.mit.edu/photonics.pdf"
   }
 }
 ```
 
-**For a single URL:**
+**Response format:**
 ```json
 {
-  "tool": "firecrawl:firecrawl_scrape",
-  "arguments": {
-    "url": "https://example.mit.edu/photonics.pdf",
-    "formats": ["markdown"],
-    "onlyMainContent": true
-  }
+  "title": "Silicon Photonics for AI - MIT",
+  "text": "# Silicon Photonics for AI\n\nRecent advances in silicon photonics have enabled...",
+  "url": "https://example.mit.edu/photonics.pdf"
 }
 ```
 
-**Response format:**
+**Note:** Call `read_webpage` for each URL sequentially or in parallel. The tool automatically converts HTML to readable text and handles most page types.
+
+### Step 4: Execute Patent Queries via Google Patents MCP
+
+For each query where `query_family == "patent"`:
+
 ```json
 {
-  "success": true,
-  "data": {
-    "markdown": "# Silicon Photonics for AI\n\nRecent advances...",
-    "metadata": {
-      "title": "Silicon Photonics for AI",
-      "sourceURL": "https://example.mit.edu/photonics.pdf"
-    }
+  "tool": "google-patents:search_patents",
+  "arguments": {
+    "q": "photonic neural network accelerator",
+    "num": 10,
+    "country": "US",
+    "after": "publication:20200101",
+    "sort": "new"
   }
 }
 ```
 
-### Step 4: Execute Patent Queries via Google Patents MCP
+**IMPORTANT:** Always specify `sort: "new"` or `sort: "old"`. The default `sort: "relevance"` is NOT supported by SerpApi and will cause an error.
 
-For each query where `query_family == "patent"`:
+**Parameter notes:**
+- `q` - Search query (required). Use semicolons to separate terms: `"(photonic) OR (optical);neural network"`
+- `num` - Results per page (10-100, default: 10)
+- `country` - Filter by country codes: `"US"`, `"US,WO,EP"`
+- `after` - Date filter format: `"publication:YYYYMMDD"` or `"filing:YYYYMMDD"`
+- `before` - Date filter format: `"publication:YYYYMMDD"` or `"filing:YYYYMMDD"`
+- `status` - Filter: `"GRANT"` or `"APPLICATION"`
+- `sort` - **Must be `"new"` or `"old"`** (NOT `"relevance"`)
 
 ```json
+// CORRECT - explicit sort parameter
 {
   "tool": "google-patents:search_patents",
   "arguments": {
-    "query": "photonic neural network accelerator",
-    "num_results": 10,
-    "country": "US",
-    "after": "2020-01-01"
+    "q": "(optical computing) AND (neural network)",
+    "num": 10,
+    "country": "US,WO",
+    "after": "publication:20180101",
+    "sort": "new"
   }
 }
 ```
@@ -234,7 +236,7 @@ Transform all responses into the standard schema with sequential source IDs:
 2. Deduplicate URLs before assigning IDs
 3. Patents get `source_type: "patent"` with extra fields
 4. Web content gets `source_type: "web"`
-5. Firecrawl markdown content goes in `excerpt` (truncated if needed)
+5. The `text` from `read_webpage` goes in `excerpt` (truncated if needed)
 
 ### Step 6: Call `sequential_research_compile`
 
@@ -280,8 +282,8 @@ await fs.writeFile(`research/${slug}/raw_results.json`, JSON.stringify(rawResult
 | Step | MCP Server | Tool | Purpose |
 |------|-----------|------|---------|
 | 1 | sequential-research | `sequential_research_plan` | Generate structured query plan |
-| 2 | google-search | `search` | Get web search results |
-| 3 | firecrawl | `firecrawl_scrape` / `firecrawl_batch_scrape` | Extract full page content |
+| 2 | google-search | `google_search` | Get web search results |
+| 3 | google-search | `read_webpage` | Extract full page content |
 | 4 | google-patents | `search_patents` | Search patent database |
 | 5 | — | — | Normalize to raw_results[] |
 | 6 | sequential-research | `sequential_research_compile` | Generate report with citations |
@@ -289,31 +291,40 @@ await fs.writeFile(`research/${slug}/raw_results.json`, JSON.stringify(rawResult
 
 ---
 
-## Firecrawl Integration Details
+## Web Scraping with `read_webpage`
+
+The `google-search:read_webpage` tool provides a simple, reliable way to fetch web content without additional MCP server dependencies.
 
-### When to Use Batch vs Single Scrape
+### Usage
 
-| Scenario | Tool | Reason |
-|----------|------|--------|
-| 5+ URLs from same query | `firecrawl_batch_scrape` | Efficiency, rate limiting |
-| 1-4 URLs | `firecrawl_scrape` (multiple calls) | Lower overhead |
-| PDF documents | `firecrawl_scrape` | Better PDF handling |
-| Real-time updates needed | `firecrawl_scrape` | Immediate results |
+```json
+{
+  "tool": "google-search:read_webpage",
+  "arguments": {
+    "url": "https://example.com/article"
+  }
+}
+```
 
-### Firecrawl Options
+### Response Format
 
 ```json
 {
-  "formats": ["markdown"],        // Output format
-  "onlyMainContent": true,        // Skip headers/footers
-  "includeTags": ["article", "main", "section"],  // Focus on content
-  "excludeTags": ["nav", "footer", "aside"],      // Skip navigation
-  "waitFor": 2000,                // Wait for JS rendering (ms)
-  "timeout": 30000                // Request timeout (ms)
+  "title": "Article Title",
+  "text": "The full text content of the page...",
+  "url": "https://example.com/article"
 }
 ```
 
-### Handling Firecrawl Errors
+### Features
+
+- **Automatic HTML to text conversion** — Clean, readable output
+- **No additional setup** — Uses the same google-search MCP server
+- **Handles most page types** — HTML, some PDFs, etc.
+
+### Handling Scrape Errors
+
+If a URL fails to scrape, include it in results with a note:
 
 ```json
 {
@@ -326,13 +337,24 @@ await fs.writeFile(`research/${slug}/raw_results.json`, JSON.stringify(rawResult
       "source_type": "web",
       "title": "Page Title (scrape failed)",
       "url": "https://example.com/blocked",
-      "excerpt": "[Content unavailable - scrape blocked by robots.txt]"
+      "excerpt": "[Content unavailable - page could not be fetched]"
     }
   ],
   "execution_notes": "1 of 5 URLs failed to scrape"
 }
 ```
 
+### Parallel Execution
+
+You can call `read_webpage` for multiple URLs in parallel to improve throughput:
+
+```
+// Execute these concurrently:
+google-search:read_webpage({url: "https://site1.com/page"})
+google-search:read_webpage({url: "https://site2.com/page"})
+google-search:read_webpage({url: "https://site3.com/page"})
+```
+
 ## Citation Format Requirement
 
 Citations must be **stable** and **machine-parseable** for downstream claim-mining.

package/examples/example_workflow.md

@@ -119,15 +119,18 @@ Execute queries using appropriate MCP tools. Here's how each query maps to data
   "params": {
     "name": "search_patents",
     "arguments": {
-      "query": "photonic computing neural network inference",
-      "num_results": 5,
+      "q": "photonic computing neural network inference",
+      "num": 10,
       "country": "US",
-      "after": "2020-01-01"
+      "after": "publication:20200101",
+      "sort": "new"
     }
   }
 }
 ```
 
+**IMPORTANT:** Always use `sort: "new"` or `sort: "old"`. The default `sort: "relevance"` is NOT supported by SerpApi and will cause an error.
+
 ### Example: Query q1 via Google Search MCP
 
 ```json

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mcp-sequential-research",
-  "version": "1.0.0",
+  "version": "1.1.1",
   "description": "MCP server for sequential research planning and compilation",
   "type": "module",
   "main": "dist/index.js",