prisma-cloud-docs-mcp-server

Instructions

Implementation Reference

• server.py:191-195 (handler)

  The handler function for the 'search_prisma_docs' tool, registered with @mcp.tool(). It searches Prisma Cloud documentation using the DocumentationIndexer and returns JSON-formatted results.

  @mcp.tool()
  async def search_prisma_docs(query: str) -> str:
      """Search Prisma Cloud documentation"""
      results = await indexer.search_docs(query, site='prisma_cloud')
      return json.dumps(results, indent=2)

• src/main.py:195-199 (handler)

  Duplicate handler function for the 'search_prisma_docs' tool in the HTTP server variant, identical implementation.

  @mcp.tool()
  async def search_prisma_docs(query: str) -> str:
      """Search Prisma Cloud documentation"""
      results = await indexer.search_docs(query, site='prisma_cloud')
      return json.dumps(results, indent=2)

• server.py:104-156 (helper)

  Core helper method in DocumentationIndexer that performs the actual search: computes relevance scores for cached pages based on title and content matches, generates snippets, sorts by score, and limits to top 10 results.

  async def search_docs(self, query: str, site: str = None) -> List[Dict]:
      """Search indexed documentation"""
      if not self.cached_pages:
          return []
      
      query_lower = query.lower()
      results = []
      
      for url, page in self.cached_pages.items():
          # Filter by site if specified
          if site and page.site != site:
              continue
          
          # Calculate relevance score
          score = 0
          title_lower = page.title.lower()
          content_lower = page.content.lower()
          
          # Higher score for title matches
          if query_lower in title_lower:
              score += 10
              # Even higher for exact title matches
              if query_lower == title_lower:
                  score += 20
          
          # Score for content matches
          content_matches = content_lower.count(query_lower)
          score += content_matches * 2
          
          # Score for partial word matches in title
          query_words = query_lower.split()
          for word in query_words:
              if word in title_lower:
                  score += 5
              if word in content_lower:
                  score += 1
          
          if score > 0:
              # Extract snippet around first match
              snippet = self._extract_snippet(page.content, query, max_length=200)
              
              results.append({
                  'title': page.title,
                  'url': page.url,
                  'site': page.site,
                  'snippet': snippet,
                  'score': score
              })
      
      # Sort by relevance score (highest first) and limit results
      results.sort(key=lambda x: x['score'], reverse=True)
      return results[:10]

• server.py:13-24 (helper)

  Dataclass used to cache indexed documentation pages with expiration logic, used by the indexer for search.

  @dataclass
  class CachedPage:
      title: str
      content: str
      url: str
      site: str
      timestamp: float
      ttl: float = 3600  # 1 hour default TTL
      
      @property
      def is_expired(self) -> bool:
          return time.time() > self.timestamp + self.ttl

• server.py:191-195 (registration)

  Registration of the tool via @mcp.tool() decorator in FastMCP framework.

  @mcp.tool()
  async def search_prisma_docs(query: str) -> str:
      """Search Prisma Cloud documentation"""
      results = await indexer.search_docs(query, site='prisma_cloud')
      return json.dumps(results, indent=2)